[C11/C2x] Change the behavior of the implicit function declaration warning

[aaron.ballman]

Summary

C89 had a questionable feature where the compiler would implicitly declare a function that the user called but was never previously declared. The resulting function would be globally declared as extern int func(); -- a function without a prototype which accepts zero or more arguments.

C99 removed support for this questionable feature due to severe security concerns. However, there was no deprecation period; C89 had the feature, C99 didn't. So Clang (and GCC) both supported the functionality as an extension in C99 and later modes.

C2x no longer supports that function signature as it now requires all functions to have a prototype, and given the known security issues with the feature, continuing to support it as an extension is not tenable.

This patch changes the diagnostic behavior for the -Wimplicit-function-declaration warning group depending on the language mode in effect. We continue to warn by default in C89 mode (due to the feature being dangerous to use), and we continue to warn by default as an extension in C99 mode (due to the lack of a deprecation period). However, because this feature will not be supported in C2x mode, the security concerns with the feature, and the trivial workaround for users (declare the function), we now default the extension warning to an error in C11 and C17 mode. This still gives users an easy workaround if they are extensively using the extension in those modes (they can disable the warning or use -Wno-error to downgrade the error), but the new diagnostic makes it more clear that this feature is not supported in C2x and should be avoided. In C2x mode, we no longer allow an implicit function to be defined and treat the situation the same as any other lookup failure.

[xbolva00]

clang/docs/ReleaseNotes.rst

114

I would mention explicitly -Wno-error= implicit-function-declaration as -Wno-error is a big hammer.

[aaron.ballman]

Fix clang-tools-extra test caught by precommit CI.

[erichkeane]

I don't have any real comments (the changes are quite trivial), and I really like the approach here, I think it is more than generous. I see the transition of
C89/C99: Warn
C11/C17: Warn-as-default-error
C2x: Hard error

as a really gentle and explainable behavior, its just a shame it took us this long to do so.

I DO see that there are a ton of AArch64 Intrin tests that seem to have the same problems here (in Pre-commit-CI), that will need this change too, though they are questionably written. I might suggest just adding the -std=c99 flag to all of those. Attn: @paulwalker-arm @MattDevereau @david-arm @sdesmalen

clang/docs/ReleaseNotes.rst

114

I think this is a valuable change, as suggested by @xbolva00

[aaron.ballman]

In D122983#3426261, @erichkeane wrote:

I don't have any real comments (the changes are quite trivial), and I really like the approach here, I think it is more than generous. I see the transition of
C89/C99: Warn
C11/C17: Warn-as-default-error
C2x: Hard error

as a really gentle and explainable behavior, its just a shame it took us this long to do so.

Thanks!

I DO see that there are a ton of AArch64 Intrin tests that seem to have the same problems here (in Pre-commit-CI), that will need this change too, though they are questionably written. I might suggest just adding the -std=c99 flag to all of those. Attn: @paulwalker-arm @MattDevereau @david-arm @sdesmalen

Yeah, those tests seem to be overly-constraining. There's no reason to be validating whether there's an implicit function declaration warning in a *codegen* test. I will change all of those AAarch64 tests to require -std=c99 explicitly whenever possible, remove the -verify flag because there's no reason for a codegen test to verify diagnostic behavior that isn't generated by the CodeGen library, and remove the // expected-warning comments. I plan to do that as an NFC change that I'll land outside of this patch, unless any of the AArch64 folks speak up pretty quickly.

clang/docs/ReleaseNotes.rst

114

Agreed!

[paulwalker-arm]

Please consider this "AArch64 folks speaking up". What are your plans here exactly? I have no issue with adding -std=c99 to the RUN lines, but "remove the // expected-warning comments" sounds like a significant loss of test coverage.

From reading the intent of the patch is it the case that some other diagnostic needs to be caught? Possibly an error diagnostic is we don't add -std=c99 and stick with the default std?

[aaron.ballman]

In D122983#3426534, @paulwalker-arm wrote:

Please consider this "AArch64 folks speaking up". What are your plans here exactly? I have no issue with adding -std=c99 to the RUN lines, but "remove the // expected-warning comments" sounds like a significant loss of test coverage.

CodeGen tests don't typically add -verify unless the diagnostic is generated by the CodeGen library, so if removing the expected-warning lines loses test coverage, that speaks to a lack of test coverage in Sema. But what, exactly, are these tests trying to validate? We have Sema tests which test the behavior of "do we want to diagnose this as an implicit declaration or not", so I don't see these lines adding any value to the tests.

[paulwalker-arm]

The tests verify a set of builtins do not exist when the associated feature flag is not present. They sit within CodeGen because the tests were plentiful and it did not seem worth duplicating them.

[hubert.reinterpretcast]

In D122983#3426450, @aaron.ballman wrote:

Yeah, those tests seem to be overly-constraining. There's no reason to be validating whether there's an implicit function declaration warning in a *codegen* test. I will change all of those AAarch64 tests to require -std=c99 explicitly whenever possible, remove the -verify flag because there's no reason for a codegen test to verify diagnostic behavior that isn't generated by the CodeGen library, and remove the // expected-warning comments. I plan to do that as an NFC change that I'll land outside of this patch, unless any of the AArch64 folks speak up pretty quickly.

Maybe a smaller hammer can be used here, e.g., -Wno-error=implicit-function-declaration? The use of the -verify flag in a CodeGen test to validate that the test is written as "cleanly" as intended is something that I am sympathetic to.

Edit: That will still lose test coverage because the warnings are checked here for a different reason.

Edit 2: It doesn't lose test coverage because the messages will still be there.

[hubert.reinterpretcast]

In D122983#3426541, @paulwalker-arm wrote:

The tests verify a set of builtins do not exist when the associated feature flag is not present. They sit within CodeGen because the tests were plentiful and it did not seem worth duplicating them.

This sounds like a situation where the test files should have more comments near the RUN lines (if the situation is not resolved via a different solution).

[aaron.ballman]

In D122983#3426541, @paulwalker-arm wrote:

The tests verify a set of builtins do not exist when the associated feature flag is not present. They sit within CodeGen because the tests were plentiful and it did not seem worth duplicating them.

You don't need to duplicate hundreds of tests to have appropriate Sema tests for this functionality. You could have added a single test file to test/Sema that tests each of the builtins you care about the semantic diagnostic behavior of. Doing this from codegen on a hundreds of individual tests that don't run everywhrere is disruptive because it makes modifying diagnostic behavior considerably harder than it should be. We separate codegen and sema tests for a reason.

I'll leave the test coverage but add -std=c99 to the verify lines because that will unblock this patch. However, I'd still appreciate if the AArch64 maintainers would clean up these tests in a more appropriate fashion, because adding the -std= line also loses test coverage (especially when we switch the default language mode). This will need to happen sooner rather than later (relatively speaking; nothing is on fire requiring immediate attention) -- I expect we'll bump clang's default from C17 to C23 sometime in the next few years, at which point all of these tests will fail to compile because there will be no support for implicit function declarations whatsoever.

[paulwalker-arm]

Thanks for this. I can see about cleaning up the tests but I'm still not sure what the advantage is. The affected RUN lines are already -fsyntax-only tests. Is it about where the test files live or is there something else I should be considering? The benefit of the current tests is that it's easy to spot holes as a single function tests both requirements. My fear is that separating the tests based on Sema/CodeGen could mean we'll miss something and never know.

[erichkeane]

In D122983#3426661, @paulwalker-arm wrote:

Thanks for this. I can see about cleaning up the tests but I'm still not sure what the advantage is. The affected RUN lines are already -fsyntax-only tests. Is it about where the test files live or is there something else I should be considering? The benefit of the current tests is that it's easy to spot holes as a single function tests both requirements. My fear is that separating the tests based on Sema/CodeGen could mean we'll miss something and never know.

There are two big problems with the way these are split up. First, the 'list' of these functions that you expect to not be declared under a flag don't exist all in 1 place, and end up being difficult to follow/track down as they are spread out as much as they are. Second, they are in tests that have a "REQUIRES" clause in it, which causes them to not be caught in many configurations. We typically avoid doing -verify in CodeGen (unless the diagnostic is ACTUALLY in CodeGen) as a matter of business. The diagnostic you're using to validate these is likely a poor choice (as it is a non-standard extension based warning) though, which is what makes this so difficult on folks trying to modernize Clang to follow standards. IMO, if these had been C++ tests (which would have 'not found' failures), or validated the existence of these in some other mechanism, we likely would have never noticed other than via an audit.

[paulwalker-arm]

Ok, message understood. I'll try and get these cleaned up, I cannot say precisely when but will push for before we branch for 15 if that sounds sensible.

[aaron.ballman]

In D122983#3426661, @paulwalker-arm wrote:

Thanks for this.

Any time, thank you for speaking up when I was about to drop test coverage by accident!

In D122983#3426716, @erichkeane wrote:

In D122983#3426661, @paulwalker-arm wrote:

Thanks for this. I can see about cleaning up the tests but I'm still not sure what the advantage is. The affected RUN lines are already -fsyntax-only tests. Is it about where the test files live or is there something else I should be considering? The benefit of the current tests is that it's easy to spot holes as a single function tests both requirements. My fear is that separating the tests based on Sema/CodeGen could mean we'll miss something and never know.

There are two big problems with the way these are split up. First, the 'list' of these functions that you expect to not be declared under a flag don't exist all in 1 place, and end up being difficult to follow/track down as they are spread out as much as they are. Second, they are in tests that have a "REQUIRES" clause in it, which causes them to not be caught in many configurations. We typically avoid doing -verify in CodeGen (unless the diagnostic is ACTUALLY in CodeGen) as a matter of business. The diagnostic you're using to validate these is likely a poor choice (as it is a non-standard extension based warning) though, which is what makes this so difficult on folks trying to modernize Clang to follow standards. IMO, if these had been C++ tests (which would have 'not found' failures), or validated the existence of these in some other mechanism, we likely would have never noticed other than via an audit.

+1 to this.

It's also about expectations -- the test directories are set up to be largely reflective of the libraries which comprise Clang and the contents of those directories are intended to test that particular component and not others. Mixing codegen and sema tests winds up being a surprise quite often when working on diagnostics because running the *whole* test suite is expensive while running just the Sema tests is quite quick. Then we can run the whole test suite once locally when we're finished changing diagnostics, just in case we missed stuff. So already, codegen tests with diagnostic checks come as a surprise from that workflow. But as Erich points out, these particular tests don't run everywhere to begin with, so it came as a rather unpleasant surprise that there's ~200 more tests impacted by changing the diagnostic. It would have been far less painful if these were in a single file in Sema -- we'd only have to fix it once instead of ~200 times and we'd catch the diagnostic regardless of what targets are registered.

[aaron.ballman]

In D122983#3426739, @paulwalker-arm wrote:

Ok, message understood. I'll try and get these cleaned up, I cannot say precisely when but will push for before we branch for 15 if that sounds sensible.

Oops, sorry, I didn't see this came in while I was also responding. :-) That would be fantastic and sounds like a very reasonable timeframe. I don't expect this to be a Real Issue until we go to change to C23 by default, which is still likely to be a few years away. Thank you for looking into it, and thanks again for speaking up about the potential loss of test coverage!

[aaron.ballman]

I pushed 5d90004874c7b040cd43597826aabb34757d25ab to hopefully address the lion's share of the currently failing precommit CI tests. I'll address the remaining few as an update to this patch.

[aaron.ballman]

Updated the release note and hopefully addressed the remaining test failures found by precommit CI.

[aaron.ballman]

Fixed a failing test case caught by precommit CI.

[xbolva00]

Could you please check that https://github.com/llvm/llvm-test-suite is buildable with your patch?

[hubert.reinterpretcast]

In D122983#3426716, @erichkeane wrote:

We typically avoid doing -verify in CodeGen (unless the diagnostic is ACTUALLY in CodeGen) as a matter of business.

I hope that -verify and // expected-no-diagnostics in CodeGen tests is compatible with the above. I believe it is valuable to confirm that the test itself is not written problematically.

[aaron.ballman]

Fixing the clangd include fixer unit tests.

[aaron.ballman]

In D122983#3427518, @xbolva00 wrote:

Could you please check that https://github.com/llvm/llvm-test-suite is buildable with your patch?

I gave it a shot just to see, but I'm unable to build it even without my patch:

F:\source\test-suite-build>cmake --build . --clean-first
Microsoft (R) Build Engine version 16.11.2+f32259642 for .NET Framework
Copyright (C) Microsoft Corporation. All rights reserved.

F:\source\test-suite-build\test-suite.sln : Solution file error MSB5004: The solution file has two projects named "pa
thfinder".
Microsoft (R) Build Engine version 16.11.2+f32259642 for .NET Framework
Copyright (C) Microsoft Corporation. All rights reserved.

  [TEST_SUITE_HOST_CC] Compiling host source fpcmp.c
  'cc' is not recognized as an internal or external command,
  operable program or batch file.
C:\Program Files (x86)\Microsoft Visual Studio\2019\Professional\MSBuild\Microsoft\VC\v160\Microsoft.CppCommon.target
s(241,5): error MSB8066: Custom build for 'F:\source\test-suite-build\CMakeFiles\a30a18e8f3bb5de504f740ee3822d8a0\fpc
mp.c.o.rule;F:\source\test-suite-build\CMakeFiles\a30a18e8f3bb5de504f740ee3822d8a0\fpcmp.rule;F:\source\test-suite-bu
ild\CMakeFiles\22099cfe59041de58024b8c82a571139\build-fpcmp.rule' exited with code 9009. [F:\source\test-suite-build\
tools\build-fpcmp.vcxproj]
  [TEST_SUITE_HOST_CC] Compiling host source timeit.c
  'cc' is not recognized as an internal or external command,
  operable program or batch file.
C:\Program Files (x86)\Microsoft Visual Studio\2019\Professional\MSBuild\Microsoft\VC\v160\Microsoft.CppCommon.target
s(241,5): error MSB8066: Custom build for 'F:\source\test-suite-build\CMakeFiles\a30a18e8f3bb5de504f740ee3822d8a0\tim
eit.c.o.rule;F:\source\test-suite-build\CMakeFiles\a30a18e8f3bb5de504f740ee3822d8a0\timeit.rule;F:\source\test-suite-
build\CMakeFiles\22099cfe59041de58024b8c82a571139\build-timeit.rule' exited with code 9009. [F:\source\test-suite-bui
ld\tools\build-timeit.vcxproj]
cl : command line warning D9025: overriding '/W4' with '/w' [F:\source\test-suite-build\MicroBenchmarks\libs\benchmar
k\third_party\googletest\build\googlemock\gmock_main.vcxproj]
cl : command line error D8021: invalid numeric argument '/Werror=date-time' [F:\source\test-suite-build\MicroBenchmar
ks\libs\benchmark\third_party\googletest\build\googlemock\gmock_main.vcxproj]
cl : command line warning D9025: overriding '/W3' with '/w' [F:\source\test-suite-build\tools\fpcmp-target.vcxproj]
cl : command line error D8021: invalid numeric argument '/Werror=date-time' [F:\source\test-suite-build\tools\fpcmp-t
arget.vcxproj]
  Generating sqlite test inputs
  'TCL_TCLSH-NOTFOUND' is not recognized as an internal or external command,
  operable program or batch file.
C:\Program Files (x86)\Microsoft Visual Studio\2019\Professional\MSBuild\Microsoft\VC\v160\Microsoft.CppCommon.target
s(241,5): error MSB8066: Custom build for 'F:\source\test-suite-build\CMakeFiles\47602658443eaf701f82747ba4e9979c\tes
t15.sql.rule' exited with code 9009. [F:\source\test-suite-build\MultiSource\Applications\sqlite3\sqlite_input.vcxpro
j]
cl : command line warning D9025: overriding '/W3' with '/w' [F:\source\test-suite-build\tools\timeit-target.vcxproj]
cl : command line error D8021: invalid numeric argument '/Werror=date-time' [F:\source\test-suite-build\tools\timeit-
target.vcxproj]

FWIW, this was how I configured, and that step did not generate any errors for me:

cmake -DCMAKE_C_COMPILER=F:\source\llvm-project\llvm\out\build\x64-Debug\bin\clang.exe -DCMAKE_CXX_COMPILER="F:\source\llvm-project\llvm\out\build\x64-Debug\bin\clang++.exe" -C..\test-suite\cmake\caches\O3.cmake -DTEST_SUITE_COLLECT_CODE_SIZE=OFF ..\test-suite

So I'm not certain what's going on there.

In D122983#3427677, @hubert.reinterpretcast wrote:

In D122983#3426716, @erichkeane wrote:

We typically avoid doing -verify in CodeGen (unless the diagnostic is ACTUALLY in CodeGen) as a matter of business.

I hope that -verify and // expected-no-diagnostics in CodeGen tests is compatible with the above. I believe it is valuable to confirm that the test itself is not written problematically.

Morally, yes, that's reasonable in CodeGen because you're ensuring you get no diagnostics. Practically, that's a convoluted, more expensive, less maintainable way to spell -Werror for the test. When diagnostics are introduced, this pattern encourages people to remove the // expected-no-diagnostics comment and start adding // expected-warning {{}} comments. Running the diagnostic verifier also slows down test execution because of the extra verification step (which adds up over thousands of tests).

[aaron.ballman]

Fixing more precommit CI failures. The ScudoWrappersCTest.MallocInfo do not appear to be caused by this patch, however.

[hubert.reinterpretcast]

In D122983#3429060, @aaron.ballman wrote:

Morally, yes, that's reasonable in CodeGen because you're ensuring you get no diagnostics. Practically, that's a convoluted, more expensive, less maintainable way to spell -Werror for the test. When diagnostics are introduced, this pattern encourages people to remove the // expected-no-diagnostics comment and start adding // expected-warning {{}} comments. Running the diagnostic verifier also slows down test execution because of the extra verification step (which adds up over thousands of tests).

Thanks for the tip. I never thought to use -Werror that way.

[aaron.ballman]

Rebase; NFC.

[erichkeane]

In D122983#3427518, @xbolva00 wrote:

Could you please check that https://github.com/llvm/llvm-test-suite is buildable with your patch?

Aaron wasn't able to get this working on his system, but I was. I ended up collecting the following errors (though, if the build ended early and my workaround of adding the flag failed, i likely missed others):

/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -D__USE_MISC -D__USE_GNU -D__USE_SVID -D__USE_XOPEN_EXTENDED -D__USE_XOPEN -Dunix -MD -MT MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o -MF MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o.d -o MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Applications/siod/slibu.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Applications/siod/slibu.c:1720:6: error: implicit declaration of function 'strptime' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
 if (strptime(get_c_string(str),get_c_string(fmt),&tm))
     ^
1 error generated.




/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -MD -MT MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o -MF MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o.d -o MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/plot2fig/fontname.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/plot2fig/fontname.c:98:18: error: implicit declaration of function 're_comp' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
  res = (char *) re_comp (s);
                 ^
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/plot2fig/fontname.c:106:7: error: implicit declaration of function 're_exec' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
  if (re_exec("typewriter")
      ^
2 errors generated.



/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG -I/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/archie-client -O3   -w -Werror=date-time -MD -MT MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o -MF MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o.d -o MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/archie-client/dirsend.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/archie-client/dirsend.c:298:17: error: implicit declaration of function 'time' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
        srand(getpid()+time(0)); /* XXX: arg ok, but not right type. */
                       ^
1 error generated.


/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -Dconst= -MD -MT MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o -MF MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o.d -o MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/MiBench/office-ispell/term.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/MiBench/office-ispell/term.c:519:9: error: implicit declaration of function 'wait' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
        while (wait (&termstat) != i)
               ^
1 error generated.


/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -w -MD -MT SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o -MF SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o.d -o SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o   -c /iusers/ekeane1/workspaces/test-suite/SingleSource/Regression/C/gcc-c-torture/execute/cmpsi-1.c
/iusers/ekeane1/workspaces/test-suite/SingleSource/Regression/C/gcc-c-torture/execute/cmpsi-1.c:4:5: error: implicit declaration of function 'dummy' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
    dummy ();
    ^
/iusers/ekeane1/workspaces/test-suite/SingleSource/Regression/C/gcc-c-torture/execute/cmpsi-1.c:15:5: error: implicit declaration of function 'dummy' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
    dummy ();
    ^
2 errors generated.

With the following 'diff', I was able to get through these:

[ekeane1@scsel-clx-24 test-suite]$ git diff
diff --git a/MultiSource/Applications/siod/CMakeLists.txt b/MultiSource/Applications/siod/CMakeLists.txt
index a562f50d..6473e22e 100644
--- a/MultiSource/Applications/siod/CMakeLists.txt
+++ b/MultiSource/Applications/siod/CMakeLists.txt
@@ -1,4 +1,4 @@
-list(APPEND CPPFLAGS -D__USE_MISC -D__USE_GNU -D__USE_SVID -D__USE_XOPEN_EXTENDED -D__USE_XOPEN -Dunix)
+list(APPEND CPPFLAGS -D__USE_MISC -D__USE_GNU -D__USE_SVID -D__USE_XOPEN_EXTENDED -D__USE_XOPEN -Dunix -Wno-implicit-function-declaration)
 list(APPEND LDFLAGS -lm)
 set(RUN_OPTIONS -v1 test.scm)
 llvm_multisource(siod)
diff --git a/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt b/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt
index 13920af8..a102311b 100644
--- a/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt
+++ b/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt
@@ -1,4 +1,4 @@
-list(APPEND CPPFLAGS -Dconst=)
+list(APPEND CPPFLAGS  -Wno-implicit-function-declaration -Dconst=)
 list(APPEND LDFLAGS -lm)
 set(RUN_OPTIONS -a -d americanmed+ < large.txt)
 llvm_multisource(office-ispell)
diff --git a/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt b/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt
index bf3b189b..f939f2a3 100644
--- a/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt
+++ b/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt
@@ -1,4 +1,5 @@
 include(CheckFunctionExists)
+list(APPEND CPPFLAGS -Wno-implicit-function-declaration)
 check_function_exists(re_comp HAVE_RE_COMP)

 if(HAVE_RE_COMP)
diff --git a/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt b/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt
index 06090ee0..c7019582 100644
--- a/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt
+++ b/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt
@@ -1,5 +1,6 @@
 include(CheckFunctionExists)
 check_function_exists(re_comp HAVE_RE_COMP)
+list(APPEND CPPFLAGS -Wno-implicit-function-declaration)
 if(HAVE_RE_COMP)
   list(APPEND LDFLAGS -lm)
   llvm_multisource(plot2fig)
diff --git a/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt b/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt
index 4daf07e7..d95836af 100644
--- a/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt
+++ b/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt
@@ -1,7 +1,7 @@
 add_subdirectory(ieee)

 # GCC C Torture Suite is conventionally run without warnings
-list(APPEND CFLAGS "-w")
+list(APPEND CFLAGS  -Wno-implicit-function-declaration "-w")

 set(TestsToSkip)

[ekeane1@scsel-clx-24 test-suite]$

I note that we likely cannot add those until the flag is added to 'clang', so either these changes need to happen at the same time, or we need to add the clang-warning-flag WITHOUT this patch's changes, THEN add these changes to the cmake lists, THEN add the rest of this patch. (Or some variety of other things).

[aaron.ballman]

In D122983#3430706, @erichkeane wrote:

In D122983#3427518, @xbolva00 wrote:

Could you please check that https://github.com/llvm/llvm-test-suite is buildable with your patch?

Aaron wasn't able to get this working on his system, but I was. I ended up collecting the following errors (though, if the build ended early and my workaround of adding the flag failed, i likely missed others):

Thank you for the help with this, Erich!

/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -D__USE_MISC -D__USE_GNU -D__USE_SVID -D__USE_XOPEN_EXTENDED -D__USE_XOPEN -Dunix -MD -MT MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o -MF MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o.d -o MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Applications/siod/slibu.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Applications/siod/slibu.c:1720:6: error: implicit declaration of function 'strptime' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
 if (strptime(get_c_string(str),get_c_string(fmt),&tm))
     ^
1 error generated.




/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -MD -MT MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o -MF MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o.d -o MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/plot2fig/fontname.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/plot2fig/fontname.c:98:18: error: implicit declaration of function 're_comp' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
  res = (char *) re_comp (s);
                 ^
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/plot2fig/fontname.c:106:7: error: implicit declaration of function 're_exec' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
  if (re_exec("typewriter")
      ^
2 errors generated.



/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG -I/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/archie-client -O3   -w -Werror=date-time -MD -MT MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o -MF MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o.d -o MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/archie-client/dirsend.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/archie-client/dirsend.c:298:17: error: implicit declaration of function 'time' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
        srand(getpid()+time(0)); /* XXX: arg ok, but not right type. */
                       ^
1 error generated.


/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -Dconst= -MD -MT MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o -MF MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o.d -o MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/MiBench/office-ispell/term.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/MiBench/office-ispell/term.c:519:9: error: implicit declaration of function 'wait' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
        while (wait (&termstat) != i)
               ^
1 error generated.


/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -w -MD -MT SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o -MF SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o.d -o SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o   -c /iusers/ekeane1/workspaces/test-suite/SingleSource/Regression/C/gcc-c-torture/execute/cmpsi-1.c
/iusers/ekeane1/workspaces/test-suite/SingleSource/Regression/C/gcc-c-torture/execute/cmpsi-1.c:4:5: error: implicit declaration of function 'dummy' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
    dummy ();
    ^
/iusers/ekeane1/workspaces/test-suite/SingleSource/Regression/C/gcc-c-torture/execute/cmpsi-1.c:15:5: error: implicit declaration of function 'dummy' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
    dummy ();
    ^
2 errors generated.

With the following 'diff', I was able to get through these:

[ekeane1@scsel-clx-24 test-suite]$ git diff
diff --git a/MultiSource/Applications/siod/CMakeLists.txt b/MultiSource/Applications/siod/CMakeLists.txt
index a562f50d..6473e22e 100644
--- a/MultiSource/Applications/siod/CMakeLists.txt
+++ b/MultiSource/Applications/siod/CMakeLists.txt
@@ -1,4 +1,4 @@
-list(APPEND CPPFLAGS -D__USE_MISC -D__USE_GNU -D__USE_SVID -D__USE_XOPEN_EXTENDED -D__USE_XOPEN -Dunix)
+list(APPEND CPPFLAGS -D__USE_MISC -D__USE_GNU -D__USE_SVID -D__USE_XOPEN_EXTENDED -D__USE_XOPEN -Dunix -Wno-implicit-function-declaration)
 list(APPEND LDFLAGS -lm)
 set(RUN_OPTIONS -v1 test.scm)
 llvm_multisource(siod)
diff --git a/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt b/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt
index 13920af8..a102311b 100644
--- a/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt
+++ b/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt
@@ -1,4 +1,4 @@
-list(APPEND CPPFLAGS -Dconst=)
+list(APPEND CPPFLAGS  -Wno-implicit-function-declaration -Dconst=)
 list(APPEND LDFLAGS -lm)
 set(RUN_OPTIONS -a -d americanmed+ < large.txt)
 llvm_multisource(office-ispell)
diff --git a/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt b/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt
index bf3b189b..f939f2a3 100644
--- a/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt
+++ b/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt
@@ -1,4 +1,5 @@
 include(CheckFunctionExists)
+list(APPEND CPPFLAGS -Wno-implicit-function-declaration)
 check_function_exists(re_comp HAVE_RE_COMP)

 if(HAVE_RE_COMP)
diff --git a/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt b/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt
index 06090ee0..c7019582 100644
--- a/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt
+++ b/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt
@@ -1,5 +1,6 @@
 include(CheckFunctionExists)
 check_function_exists(re_comp HAVE_RE_COMP)
+list(APPEND CPPFLAGS -Wno-implicit-function-declaration)
 if(HAVE_RE_COMP)
   list(APPEND LDFLAGS -lm)
   llvm_multisource(plot2fig)
diff --git a/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt b/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt
index 4daf07e7..d95836af 100644
--- a/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt
+++ b/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt
@@ -1,7 +1,7 @@
 add_subdirectory(ieee)

 # GCC C Torture Suite is conventionally run without warnings
-list(APPEND CFLAGS "-w")
+list(APPEND CFLAGS  -Wno-implicit-function-declaration "-w")

 set(TestsToSkip)

[ekeane1@scsel-clx-24 test-suite]$

I note that we likely cannot add those until the flag is added to 'clang', so either these changes need to happen at the same time, or we need to add the clang-warning-flag WITHOUT this patch's changes, THEN add these changes to the cmake lists, THEN add the rest of this patch. (Or some variety of other things).

The -Wimplicit-function-declaration already exists today, so I believe we can land the changes you have above to the test suite already today independent of this patch. Would you be interested in being nerd-sniped into doing that @erichkeane since you can actually run the tests? (I'm happy to support you in the work however I can, of course.)

[erichkeane]

In D122983#3433685, @aaron.ballman wrote:

In D122983#3430706, @erichkeane wrote:

In D122983#3427518, @xbolva00 wrote:

Could you please check that https://github.com/llvm/llvm-test-suite is buildable with your patch?

Aaron wasn't able to get this working on his system, but I was. I ended up collecting the following errors (though, if the build ended early and my workaround of adding the flag failed, i likely missed others):

Thank you for the help with this, Erich!

/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -D__USE_MISC -D__USE_GNU -D__USE_SVID -D__USE_XOPEN_EXTENDED -D__USE_XOPEN -Dunix -MD -MT MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o -MF MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o.d -o MultiSource/Applications/siod/CMakeFiles/siod.dir/slibu.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Applications/siod/slibu.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Applications/siod/slibu.c:1720:6: error: implicit declaration of function 'strptime' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
 if (strptime(get_c_string(str),get_c_string(fmt),&tm))
     ^
1 error generated.




/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -MD -MT MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o -MF MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o.d -o MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeFiles/plot2fig.dir/fontname.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/plot2fig/fontname.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/plot2fig/fontname.c:98:18: error: implicit declaration of function 're_comp' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
  res = (char *) re_comp (s);
                 ^
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/plot2fig/fontname.c:106:7: error: implicit declaration of function 're_exec' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
  if (re_exec("typewriter")
      ^
2 errors generated.



/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG -I/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/archie-client -O3   -w -Werror=date-time -MD -MT MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o -MF MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o.d -o MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeFiles/archie.dir/dirsend.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/archie-client/dirsend.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/Prolangs-C/archie-client/dirsend.c:298:17: error: implicit declaration of function 'time' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
        srand(getpid()+time(0)); /* XXX: arg ok, but not right type. */
                       ^
1 error generated.


/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -Dconst= -MD -MT MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o -MF MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o.d -o MultiSource/Benchmarks/MiBench/office-ispell/CMakeFiles/office-ispell.dir/term.c.o   -c /iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/MiBench/office-ispell/term.c
/iusers/ekeane1/workspaces/test-suite/MultiSource/Benchmarks/MiBench/office-ispell/term.c:519:9: error: implicit declaration of function 'wait' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
        while (wait (&termstat) != i)
               ^
1 error generated.


/iusers/ekeane1/workspaces/test-suite-build/tools/timeit --summary SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o.time /iusers/ekeane1/workspaces/llvm-project2/build/bin/clang -DNDEBUG  -O3   -w -Werror=date-time -w -MD -MT SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o -MF SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o.d -o SingleSource/Regression/C/gcc-c-torture/execute/CMakeFiles/GCC-C-execute-cmpsi-1.dir/cmpsi-1.c.o   -c /iusers/ekeane1/workspaces/test-suite/SingleSource/Regression/C/gcc-c-torture/execute/cmpsi-1.c
/iusers/ekeane1/workspaces/test-suite/SingleSource/Regression/C/gcc-c-torture/execute/cmpsi-1.c:4:5: error: implicit declaration of function 'dummy' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
    dummy ();
    ^
/iusers/ekeane1/workspaces/test-suite/SingleSource/Regression/C/gcc-c-torture/execute/cmpsi-1.c:15:5: error: implicit declaration of function 'dummy' is invalid in C99 and later and unsupported in C2x [-Wimplicit-function-declaration]
    dummy ();
    ^
2 errors generated.

With the following 'diff', I was able to get through these:

[ekeane1@scsel-clx-24 test-suite]$ git diff
diff --git a/MultiSource/Applications/siod/CMakeLists.txt b/MultiSource/Applications/siod/CMakeLists.txt
index a562f50d..6473e22e 100644
--- a/MultiSource/Applications/siod/CMakeLists.txt
+++ b/MultiSource/Applications/siod/CMakeLists.txt
@@ -1,4 +1,4 @@
-list(APPEND CPPFLAGS -D__USE_MISC -D__USE_GNU -D__USE_SVID -D__USE_XOPEN_EXTENDED -D__USE_XOPEN -Dunix)
+list(APPEND CPPFLAGS -D__USE_MISC -D__USE_GNU -D__USE_SVID -D__USE_XOPEN_EXTENDED -D__USE_XOPEN -Dunix -Wno-implicit-function-declaration)
 list(APPEND LDFLAGS -lm)
 set(RUN_OPTIONS -v1 test.scm)
 llvm_multisource(siod)
diff --git a/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt b/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt
index 13920af8..a102311b 100644
--- a/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt
+++ b/MultiSource/Benchmarks/MiBench/office-ispell/CMakeLists.txt
@@ -1,4 +1,4 @@
-list(APPEND CPPFLAGS -Dconst=)
+list(APPEND CPPFLAGS  -Wno-implicit-function-declaration -Dconst=)
 list(APPEND LDFLAGS -lm)
 set(RUN_OPTIONS -a -d americanmed+ < large.txt)
 llvm_multisource(office-ispell)
diff --git a/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt b/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt
index bf3b189b..f939f2a3 100644
--- a/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt
+++ b/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt
@@ -1,4 +1,5 @@
 include(CheckFunctionExists)
+list(APPEND CPPFLAGS -Wno-implicit-function-declaration)
 check_function_exists(re_comp HAVE_RE_COMP)

 if(HAVE_RE_COMP)
diff --git a/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt b/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt
index 06090ee0..c7019582 100644
--- a/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt
+++ b/MultiSource/Benchmarks/Prolangs-C/plot2fig/CMakeLists.txt
@@ -1,5 +1,6 @@
 include(CheckFunctionExists)
 check_function_exists(re_comp HAVE_RE_COMP)
+list(APPEND CPPFLAGS -Wno-implicit-function-declaration)
 if(HAVE_RE_COMP)
   list(APPEND LDFLAGS -lm)
   llvm_multisource(plot2fig)
diff --git a/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt b/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt
index 4daf07e7..d95836af 100644
--- a/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt
+++ b/SingleSource/Regression/C/gcc-c-torture/execute/CMakeLists.txt
@@ -1,7 +1,7 @@
 add_subdirectory(ieee)

 # GCC C Torture Suite is conventionally run without warnings
-list(APPEND CFLAGS "-w")
+list(APPEND CFLAGS  -Wno-implicit-function-declaration "-w")

 set(TestsToSkip)

[ekeane1@scsel-clx-24 test-suite]$

I note that we likely cannot add those until the flag is added to 'clang', so either these changes need to happen at the same time, or we need to add the clang-warning-flag WITHOUT this patch's changes, THEN add these changes to the cmake lists, THEN add the rest of this patch. (Or some variety of other things).

The -Wimplicit-function-declaration already exists today, so I believe we can land the changes you have above to the test suite already today independent of this patch. Would you be interested in being nerd-sniped into doing that @erichkeane since you can actually run the tests? (I'm happy to support you in the work however I can, of course.)

*GRUMBLE GRUMBLE GRUMBLE*.
Sure, I can get on that in the next day or two.

[aaron.ballman]

In D122983#3433691, @erichkeane wrote:

In D122983#3433685, @aaron.ballman wrote:

The -Wimplicit-function-declaration already exists today, so I believe we can land the changes you have above to the test suite already today independent of this patch. Would you be interested in being nerd-sniped into doing that @erichkeane since you can actually run the tests? (I'm happy to support you in the work however I can, of course.)

*GRUMBLE GRUMBLE GRUMBLE*.
Sure, I can get on that in the next day or two.

:-D Thanks, I owe you Yet Another Beer!

[erichkeane]

Up for review: https://reviews.llvm.org/D123241

[aaron.ballman]

In D122983#3433919, @erichkeane wrote:

Up for review: https://reviews.llvm.org/D123241

Thank you @erichkeane! That landed in a87f4e4f9808a397dc4c575013142c9eac0b7eba, so it should no longer be blocking this patch.

[aaron.ballman]

Rebased

[jyknight]

Looks reasonable overall.

I've added a few specific comments to some tests, but in general, I think we should avoid adding -std=c99 to test-cases to workaround implicit decl issues, unless:

• the test is explicitly testing the behavior of implicit decls (potentially in interaction with some other feature).
• it's a regression test, with some particular reduced/artificial inputs.
• the ARM codegen tests, for now, on the assumption the arm maintainers will address it separately.

clang-tools-extra/test/clang-tidy/checkers/bugprone-easily-swappable-parameters-implicits.c

76

This is just a typo, you won't need the -std=c99 on this test if you fix pointeeConverison -> pointeeConversion

clang/docs/ReleaseNotes.rst

113–118

Some wording suggestions here, to remove parenthetical, and more explicitly note that the options have no effect in C2x.

clang/include/clang/Basic/DiagnosticSemaKinds.td

423–426

I think it'd be good to use the same message for both variants. (Even if you're building in c99, might as well note it's gone in C2x).

clang/test/CodeGen/PowerPC/builtins-ppc-p8vector.c

4

This test-run is already expected to fail, so the -std=c99 seems like it shouldn't be necessary. I think just change the CHECK-PPC lines looking for "warning: implicit declaration" to look for "error: implicit declaration" instead.

clang/test/CodeGen/X86/bmi2-builtins.c

1

That this is needed looks like a test bug here -- it should be testing _mulx_u64 on x86-64 and _mulx_u32 on i386.

clang/test/CodeGen/misaligned-param.c

1

Declare int bar(struct s*, struct s*); instead

clang/test/Headers/arm-cmse-header-ns.c

1

Maybe better to edit the CHECK-c lines to expect an error?

clang/test/Modules/modulemap-locations.m

2

Change expected-warnings to expected-error instead of -std=c99?

clang/test/PCH/chain-macro-override.c

13

Change to expected-error and delete -std=c99?

clang/test/Sema/implicit-decl.c

30

Yikes! That's a particularly awful typo-fix suggestion. I assume from those messages, it decided this should be a function declaration instead of a call, and is parsing as a function declaration?

I'd say that's unrelated to this change, EXCEPT, that doesn't happen in C++ mode. So I'm guessing there's some weird interaction here that triggers ONLY in the C mode of the parser, which causes it to prefer to parse unknown identifiers as a type name instead of a variable/function name (on the assumption that implicit function decls exist, and would've triggered already, if relevant, perhaps?).

clang/test/Sema/implicit-ms-builtin-decl.c

41

switch to expected-error and remove -std=c99?

clang/test/VFS/module_missing_vfs.m

2

I can't quite follow what this test is doing, but it looks like it does intend to have a prototype available?

[aaron.ballman]

Thank you for the thorough double-check on the test changes, @jyknight! I was able to remove most of the ones you called out (I commented where I couldn't) plus a few more -std=c99 from RUN lines, but there are quite a few left because it's entirely unclear what the test is actually testing (especially in CodeGen tests -- we have these RUN lines but they don't check anything, so I'm assuming these are "don't crash" tests reduced from user code, so I left the -std=c99 for them).

[aaron.ballman]

clang-tools-extra/test/clang-tidy/checkers/bugprone-easily-swappable-parameters-implicits.c

76

Good catch, I'll fix this up in an NFC change.

clang/include/clang/Basic/DiagnosticSemaKinds.td

423–426

Heh, I waffled on doing exactly that, so I'm glad for this feedback!

clang/test/CodeGen/misaligned-param.c

1

Good catch; I think I was worried this was intending to test the function call, which is why I added the -std=c89.

clang/test/Sema/implicit-decl.c

30

Yikes! That's a particularly awful typo-fix suggestion. I assume from those messages, it decided this should be a function declaration instead of a call, and is parsing as a function declaration?

That's correct.

I'd say that's unrelated to this change, EXCEPT, that doesn't happen in C++ mode. So I'm guessing there's some weird interaction here that triggers ONLY in the C mode of the parser, which causes it to prefer to parse unknown identifiers as a type name instead of a variable/function name (on the assumption that implicit function decls exist, and would've triggered already, if relevant, perhaps?).

That seems to be correct. It appears we get into this state because of implicit int. So I'd like to leave this alone for now and try to revisit later when I try to treat implicit int similar to what I've done in this patch for implicit function declarations.

clang/test/VFS/module_missing_vfs.m

2

Yeah, I struggled mightily to figure out what this test was doing and eventually gave up. I *think* the behavior of the test is that it's putting the declaration in a.h and then overlaying with an a.h (in Inputs/MissingVFS) that does not have the declaration, and demonstrating that you can still call it? But I honestly don't know.

[aaron.ballman]

Update based on review feedback.

• Removed a typo correction note so that we don't recommend a previous declaration which was implicitly declared
• Removed several -std=c99 options from various RUN lines; other updates to tests

[rsmith]

It seems surprising to me for the diagnostic to change from warn-by-default to error-by-default when changing from C99 to C11, given that the language rule did not change between C99 and C11 (as a Clang user, when changing my -std= flag, I don't want other changes to come in that are unrelated to the language mode change I requested). I think we should just make this an error by default in C99 onwards; if we're happy promoting this from warning-by-default to error-by-default for the folks using -std=c11 and later (and I think we are), then we should be happy doing the same for the -std=c99 folks too -- especially given that C17 is the default everywhere other than on PS4.

clang/lib/Sema/SemaDecl.cpp

15269

For consistency, this case should be an ExtWarn, not just a Warning, in C99 onwards (and an Error in C2x, which it will be because we don't get here in that case). But given that this is within the implementation namespace and it's currently a DefaultError, perhaps we should instead fully replace this Warning with an Error in all language modes?

[aaron.ballman]

In D122983#3450177, @rsmith wrote:

It seems surprising to me for the diagnostic to change from warn-by-default to error-by-default when changing from C99 to C11, given that the language rule did not change between C99 and C11 (as a Clang user, when changing my -std= flag, I don't want other changes to come in that are unrelated to the language mode change I requested).

FWIW, that's the approach I originally considered. I eventually discarded it as being likely too disruptive for users (with risk of pushing some people to not upgrade to Clang 15).

This situation is a somewhat unique one. C89 supported this feature and it was not obsolescent in that revision of the standard. C99 removed the feature outright. (The same thing happened to implicit int.) Because there was no deprecation period, people got caught off guard and so compilers made it a warning in C99 that also triggered in C89. As best I can tell, this is the first time the severity question has been revisited. The reason I took the approach I did is because of the lack of deprecation period and the somewhat common user practice of using an effectively-C89 code base with some C99 features (most often: ability to mix decls + code, // comments, and long long support -- aka, what MSVC considered "C" to be for many, many years).

I think we should just make this an error by default in C99 onwards; if we're happy promoting this from warning-by-default to error-by-default for the folks using -std=c11 and later (and I think we are), then we should be happy doing the same for the -std=c99 folks too -- especially given that C17 is the default everywhere other than on PS4.

I see where you're coming from, but I don't think it's practical. We've given users this as a warning in C99 and later for a *long* time (since Clang 2.6 from 2009), so I anticipate an error breaking some amount of code for people passing -std=c99 explicitly. We do still run the risk of breaking people who are on C11 or C17 (particularly, given it's the default when no -std= is provided), but that feels like a more manageable risk. Those folks are clearly interested in upgrading language modes that I don't think apply to folks specifying C99 specifically (now that it's 20+ years old).

tl;dr: I think the current approach strikes a good balance between pushing people to improve their code and giving them a reasonable upgrade path through the language standards.

[erichkeane]

FWIW, Aaron and I discussed this at length while he was preparing this patch. I think the current behavior (warn in C99, warn-as-error in C11/C17, hard error in C20) is about as aggressive as I'd want us to be with erroring without causing extensive heartache on our users. WG14 did us no favors here in 1999, and our delay on making this an error for 20 years since then has created an environment where I wish for us to be somewhat understanding of our users. If you'd asked me in 1998 of course my answer would have been different.

I think Aaron's approach provides a proper 'depreciation' period in our compiler, as best as we have the ability to do. I consider MOST projects who use C99 to be 'legacy' code bases with little maintenance, so introducing new errors to them would be unfortunate (even default-W-as-E). I see value to still warning in the C99 case, as it _IS_ a good improvement for anyone going through the source code trying to do small improvements, but a default-W-as-E causes greater overhead in that case.

Comparatively, I see projects using C11/C17 to be 'modern' or 'maintained' code bases, so the disable-able error seems to me to be a proper 'kick' to let them know this is something they need to change ASAP. I would be VERY against this being 'just a warning' though, as they are more simply/commonly ignored, and makes the hard-error in C20 that much more jarring.

While I agree this is quite a odd of a transition, I see it as the most-gentle we can be for our users. Anything 'more aggressive' in the C99/C11/C17 case would be an unfortunate burden, and anything 'less aggressive' would make the C20 change a further jarring/unfortunate burden.

[rsmith]

In D122983#3451920, @erichkeane wrote:

I think Aaron's approach provides a proper 'depreciation' period in our compiler, as best as we have the ability to do.

We've been warning by default for a decade that this code is not valid in C99; that was our deprecation period. If the aim is to provide a deprecation period, the end goal should be that in some future Clang version we complete the transition and change the C99 default to reject too. Otherwise, that's not a deprecation period, that's a permanent language extension in our C99 mode -- and it seems capricious to provide that extension by default in C99 mode but not C11 / C17 mode. Given that this extension is, well, bad, I think we shouldn't be providing it by default anywhere. It's not hard for people to turn the warning flag off, and people intentionally using this in C99 onwards probably already have done so.

[aaron.ballman]

In D122983#3452994, @rsmith wrote:

In D122983#3451920, @erichkeane wrote:

I think Aaron's approach provides a proper 'depreciation' period in our compiler, as best as we have the ability to do.

We've been warning by default for a decade that this code is not valid in C99; that was our deprecation period.

I thought about this perspective a bunch overnight and I think I agree that this is reasonable given that the warning explicitly calls the code out as being *invalid*.

If the aim is to provide a deprecation period, the end goal should be that in some future Clang version we complete the transition and change the C99 default to reject too. Otherwise, that's not a deprecation period, that's a permanent language extension in our C99 mode -- and it seems capricious to provide that extension by default in C99 mode but not C11 / C17 mode. Given that this extension is, well, bad, I think we shouldn't be providing it by default anywhere. It's not hard for people to turn the warning flag off, and people intentionally using this in C99 onwards probably already have done so.

FWIW, I misunderstood your original suggestion:

I think we should just make this an error by default in C99 onwards;

I thought you meant this should be an *error* (not a warning defaulting to an error) in C99 and later. That would be too aggressive of an approach for me to support because it gives no upgrade path to users to easily get onto Clang 15. Now that I understand you better, I like your approach and rationale.

[aaron.ballman]

In D122983#3452994, @rsmith wrote:

I think we should just make this an error by default in C99 onwards;

Out of curiosity -- do you think we should remove the DefaultIgnore in C89 mode so that we warn by default there (perhaps as a follow up)? Also, I presume you expect diag::ext_implicit_lib_function_decl to behave the same way (warn in C89, warn-as-err in C99 and up) as part of this patch?

[jyknight]

In D122983#3454406, @aaron.ballman wrote:

In D122983#3452994, @rsmith wrote:

I think we should just make this an error by default in C99 onwards;

Out of curiosity -- do you think we should remove the DefaultIgnore in C89 mode so that we warn by default there (perhaps as a follow up)? Also, I presume you expect diag::ext_implicit_lib_function_decl to behave the same way (warn in C89, warn-as-err in C99 and up) as part of this patch?

I'm not sure what purpose it'd serve to change -std=c89 to be more strict at this point. It's not the default compilation mode, and the code is actually valid under that standard. IMO, adding such a on-by-default warning there would only serve to annoy folks explicitly trying to build super-old code with a super-old standards version.

[aaron.ballman]

In D122983#3454494, @jyknight wrote:

In D122983#3454406, @aaron.ballman wrote:

In D122983#3452994, @rsmith wrote:

I think we should just make this an error by default in C99 onwards;

Out of curiosity -- do you think we should remove the DefaultIgnore in C89 mode so that we warn by default there (perhaps as a follow up)? Also, I presume you expect diag::ext_implicit_lib_function_decl to behave the same way (warn in C89, warn-as-err in C99 and up) as part of this patch?

I'm not sure what purpose it'd serve to change -std=c89 to be more strict at this point. It's not the default compilation mode, and the code is actually valid under that standard. IMO, adding such a on-by-default warning there would only serve to annoy folks explicitly trying to build super-old code with a super-old standards version.

Yeah, I was waffling on that one -- my thinking was that it would at least warn users that they're using something dangerous and we're going to break them if they attempt to upgrade (aka, we treat it as almost-deprecated in C89). But I'm fine leaving that one as DefaultIgnore too; as you say, if someone is in that mode explicitly, this IS a "feature" of that language.

I feel a bit more strongly that implicit builtins should be handled the same as other implicit functions (in terms of DefaultError behavior), unless there's some scenarios I'm not thinking of.

[rsmith]

In D122983#3454508, @aaron.ballman wrote:

In D122983#3454494, @jyknight wrote:

In D122983#3454406, @aaron.ballman wrote:

Out of curiosity -- do you think we should remove the DefaultIgnore in C89 mode so that we warn by default there (perhaps as a follow up)? Also, I presume you expect diag::ext_implicit_lib_function_decl to behave the same way (warn in C89, warn-as-err in C99 and up) as part of this patch?

I think it would make sense to include the C89 warning in -Wc99-compat, but I agree with James that it doesn't seem worthwhile to enable it by default.

I feel a bit more strongly that implicit builtins should be handled the same as other implicit functions (in terms of DefaultError behavior), unless there's some scenarios I'm not thinking of.

I'm not sure which case you mean by "implicit builtins". Two cases seem relevant:

• ext_implicit_lib_function_decl, for things like using strlen without including its header; I think that should behave the same as using any other undeclared function name (error by default in C99 onwards, error with no warning flag in C2x onwards even though we would synthesize the correct prototype).
• warn_builtin_unknown, for things like using __builtin_no_such_builtin_exists() (a function name starting __builtin_ that is not a known builtin and hasn't been declared), which is currently a warning-as-error in all language modes; I think that should stay an error by default in all modes, and I'd be happy to see it turn into a harder error (with no warning flag), whether that happens only in C2x onwards, in C99 onwards, or in all modes.

[aaron.ballman]

In D122983#3454632, @rsmith wrote:

In D122983#3454508, @aaron.ballman wrote:

In D122983#3454494, @jyknight wrote:

In D122983#3454406, @aaron.ballman wrote:

Out of curiosity -- do you think we should remove the DefaultIgnore in C89 mode so that we warn by default there (perhaps as a follow up)? Also, I presume you expect diag::ext_implicit_lib_function_decl to behave the same way (warn in C89, warn-as-err in C99 and up) as part of this patch?

I think it would make sense to include the C89 warning in -Wc99-compat, but I agree with James that it doesn't seem worthwhile to enable it by default.

Ah, that's a good idea for a follow-up, thanks!

I feel a bit more strongly that implicit builtins should be handled the same as other implicit functions (in terms of DefaultError behavior), unless there's some scenarios I'm not thinking of.

I'm not sure which case you mean by "implicit builtins". Two cases seem relevant:

• ext_implicit_lib_function_decl, for things like using strlen without including its header; I think that should behave the same as using any other undeclared function name (error by default in C99 onwards, error with no warning flag in C2x onwards even though we would synthesize the correct prototype).

Great, this was the one I was talking about, and that's the approach I'd like to take with it. I'll have that in the next revision.

• warn_builtin_unknown, for things like using __builtin_no_such_builtin_exists() (a function name starting __builtin_ that is not a known builtin and hasn't been declared), which is currently a warning-as-error in all language modes; I think that should stay an error by default in all modes, and I'd be happy to see it turn into a harder error (with no warning flag), whether that happens only in C2x onwards, in C99 onwards, or in all modes.

Okay, I hadn't touched that one because it was already default error. I'm happy to try to turn it into a hard error in a followup patch.

Thanks for the feedback!

[aaron.ballman]

Updated based on review feedback.

• The warning now defaults to an error in C99 and up.
• The implicit builtin decl warning now behaves the same as the regular implicit function decl warning

This patch is significantly larger because there were ~250 new test failures. Most of them are from the ARM codegen tests using sema diagnostics to test codegen behavior.

[aaron.ballman]

Rebased, no changes.

[aaron.ballman]

Fixed test failures in compiler-rt and clangd found by precommit CI.

[rsmith]

LGTM; two non-blocking comments and a question.

clang/include/clang/Basic/DiagnosticSemaKinds.td

422–427

The context for someone reading this diagnostic is that they're building in C99 / C11 / C17 mode, used an undeclared function, and (by default) saw this error. In that context, I think it may not be especially relevant to them that Clang would not even allow the error to be disabled if they built in C2x mode, so I'd be inclined to remove the "and unsupported in C2x" part from this diagnostic. Also, they probably didn't *intend* to implicitly declare a function, so perhaps we could replace this diagnostic with something less implementation-oriented and more user-oriented:

"call to undeclared function %0; ISO C99 and later do not support implicit function declarations"

(The phrasing "ISO C99 and later do not support" is what we usually use when we want to imply "but perhaps Clang does support" -- keeping in mind that this diagnostic might appear as either an error or a warning, and we want phrasing that works both ways.)

That said: this is orthogonal to the changes you're making here, so don't consider this comment to be blocking.

702–704

Hm, why is this one an ExtWarn? Are we really allowed to reject implicit declarations of library functions (in -std=c89 -pedantic-errors mode)?

clang/lib/Sema/SemaDecl.cpp

15319–15321

Should we even be calling this function in OpenCL mode? It seems a bit inconsistent that we avoid calling this in C++ and C2x mode, and that we call it but error in OpenCL mode.

Maybe there should be a function on LangOptions to ask if implicit function declarations are permitted in the current language mode, to make it easy to opt out the right cases? (Happy for this to be a follow-on change if you agree.)

[jyknight]

Looks good to me, whichever way you go regarding rsmith's comments.

[aaron.ballman]

Updated the diagnostic wording, putting it through precommit CI again to see if I missed any tests.

[aaron.ballman]

Rebased to hopefully get precommit CI to test it.

[aaron.ballman]

Fixes some failing tests found by precommit CI.

[aaron.ballman]

clang/include/clang/Basic/DiagnosticSemaKinds.td

422–427

In that context, I think it may not be especially relevant to them that Clang would not even allow the error to be disabled if they built in C2x mode, so I'd be inclined to remove the "and unsupported in C2x" part from this diagnostic.

The reason I had it worded that way was to justify why it's an error instead of a warning.

"call to undeclared function %0; ISO C99 and later do not support implicit function declarations"

I think this is shorter and equally as clear as what I had. I'll make the change and push it up again for precommit CI to run over (I would not be surprised if I missed a few places to update because of loose -verify checking behavior that will no longer match the new wording).

702–704

No, we're not. But ext_implicit_lib_function_decl was already ExtWarn; this new one is the diagnostic you get only in C99 and later.

If you'd like, I can make the C89 one into a Warning though (either now or in a follow-up). However, we should probably also think about whether we want C++ to be a default error or not. In this patch it does not default to an error.

clang/lib/Sema/SemaDecl.cpp

15319–15321

I agree that it does seem inconsistent. I can look into making that change in a follow-up.

[nemanjai]

This is still causing failures when building test-suite:
https://lab.llvm.org/buildbot/#/builders/105/builds/24292

Why don't we just turn off this warning for the entire test-suite since I don't expect we want to modify the tests (as some of them have licenses that probably don't allow us to modify them)?
We could do something like this:

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 8d76a45c7..032281ecf 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -47,6 +47,7 @@ set(LLVM_CODESIGNING_IDENTITY "" CACHE STRING
   "Sign executables and dylibs with the given identity or skip if empty (Darwin Only)")
 
 add_definitions(-DNDEBUG)
+add_definitions(-Wno-implicit-function-declaration)
 option(TEST_SUITE_SUPPRESS_WARNINGS "Suppress all warnings" ON)
 if(${TEST_SUITE_SUPPRESS_WARNINGS})
   add_definitions(-w)

[MaskRay]

In D122983#3463521, @nemanjai wrote:

This is still causing failures when building test-suite:
https://lab.llvm.org/buildbot/#/builders/105/builds/24292

Why don't we just turn off this warning for the entire test-suite since I don't expect we want to modify the tests (as some of them have licenses that probably don't allow us to modify them)?
We could do something like this:

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 8d76a45c7..032281ecf 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -47,6 +47,7 @@ set(LLVM_CODESIGNING_IDENTITY "" CACHE STRING
   "Sign executables and dylibs with the given identity or skip if empty (Darwin Only)")
 
 add_definitions(-DNDEBUG)
+add_definitions(-Wno-implicit-function-declaration)
 option(TEST_SUITE_SUPPRESS_WARNINGS "Suppress all warnings" ON)
 if(${TEST_SUITE_SUPPRESS_WARNINGS})
   add_definitions(-w)

Adding a global -Wno-implicit-function-declaration looks good. @Meinersbur

[xbolva00]

In D122983#3463521, @nemanjai wrote:

This is still causing failures when building test-suite:
https://lab.llvm.org/buildbot/#/builders/105/builds/24292

Why don't we just turn off this warning for the entire test-suite since I don't expect we want to modify the tests (as some of them have licenses that probably don't allow us to modify them)?
We could do something like this:

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 8d76a45c7..032281ecf 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -47,6 +47,7 @@ set(LLVM_CODESIGNING_IDENTITY "" CACHE STRING
   "Sign executables and dylibs with the given identity or skip if empty (Darwin Only)")
 
 add_definitions(-DNDEBUG)
+add_definitions(-Wno-implicit-function-declaration)
 option(TEST_SUITE_SUPPRESS_WARNINGS "Suppress all warnings" ON)
 if(${TEST_SUITE_SUPPRESS_WARNINGS})
   add_definitions(-w)

But your link shows failures in compiler-rt/, not in llvm test-suite

[nemanjai]

In D122983#3463569, @xbolva00 wrote:

But your link shows failures in compiler-rt/, not in llvm test-suite

It shows both. But the compiler-rt ones have subsequently been fixed while the test-suite ones are still failing as of https://lab.llvm.org/buildbot/#/builders/105/builds/24314.

[aaron.ballman]

In D122983#3463600, @nemanjai wrote:

In D122983#3463569, @xbolva00 wrote:

But your link shows failures in compiler-rt/, not in llvm test-suite

It shows both. But the compiler-rt ones have subsequently been fixed while the test-suite ones are still failing as of https://lab.llvm.org/buildbot/#/builders/105/builds/24314.

Thanks for pointing this out -- I'm baffled what's going on with that bot, because 1) other bots running the test suite are fine, and 2) the test suite was already fixed to disable the diagnostics (which is what fixed things for #1):

https://github.com/llvm/llvm-test-suite/blob/main/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt#L2
https://github.com/llvm/llvm-test-suite/blob/main/MultiSource/Benchmarks/Prolangs-C/bison/CMakeLists.txt#L1

(and so on). It's almost as if something on that particular bot is overriding the cmake flags (or perhaps using CFLAGS instead of CPPFLAGS)... Here's the command line that's failing for bison's symtab.c:

RunSafely.sh detected a failure with these command-line arguments: --show-errors -t /home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/test/sandbox/build/tools/timeit 500 /dev/null Output/symtab.llvm.o.compile /home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/stage1.install/bin/clang -I/home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/test/sandbox/build/MultiSource/Benchmarks/Prolangs-C/bison -I/home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/test/test-suite/MultiSource/Benchmarks/Prolangs-C/bison -I/home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/test/test-suite/include -I../../../../include -D_GNU_SOURCE -D__STDC_LIMIT_MACROS -DNDEBUG -O3 -ffp-contract=off -fomit-frame-pointer -c /home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/test/test-suite/MultiSource/Benchmarks/Prolangs-C/bison/symtab.c -o Output/symtab.llvm.o

Note how it's missing the warning flag from the CMakeLists.txt file.

So I have no idea what's going on (I can't even really run the tests; I'm on Windows and test-suite requires perl to run).

In D122983#3463561, @MaskRay wrote:

Adding a global -Wno-implicit-function-declaration looks good. @Meinersbur

+1 to this idea, but I'm worried it won't fix that particular bot.

[xbolva00]

In D122983#3464315, @aaron.ballman wrote:

In D122983#3463600, @nemanjai wrote:

In D122983#3463569, @xbolva00 wrote:

But your link shows failures in compiler-rt/, not in llvm test-suite

It shows both. But the compiler-rt ones have subsequently been fixed while the test-suite ones are still failing as of https://lab.llvm.org/buildbot/#/builders/105/builds/24314.

Thanks for pointing this out -- I'm baffled what's going on with that bot, because 1) other bots running the test suite are fine, and 2) the test suite was already fixed to disable the diagnostics (which is what fixed things for #1):

https://github.com/llvm/llvm-test-suite/blob/main/MultiSource/Benchmarks/Prolangs-C/archie-client/CMakeLists.txt#L2
https://github.com/llvm/llvm-test-suite/blob/main/MultiSource/Benchmarks/Prolangs-C/bison/CMakeLists.txt#L1

(and so on). It's almost as if something on that particular bot is overriding the cmake flags (or perhaps using CFLAGS instead of CPPFLAGS)... Here's the command line that's failing for bison's symtab.c:

RunSafely.sh detected a failure with these command-line arguments: --show-errors -t /home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/test/sandbox/build/tools/timeit 500 /dev/null Output/symtab.llvm.o.compile /home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/stage1.install/bin/clang -I/home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/test/sandbox/build/MultiSource/Benchmarks/Prolangs-C/bison -I/home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/test/test-suite/MultiSource/Benchmarks/Prolangs-C/bison -I/home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/test/test-suite/include -I../../../../include -D_GNU_SOURCE -D__STDC_LIMIT_MACROS -DNDEBUG -O3 -ffp-contract=off -fomit-frame-pointer -c /home/buildbots/ppc64le-clang-lnt-test/clang-ppc64le-lnt/test/test-suite/MultiSource/Benchmarks/Prolangs-C/bison/symtab.c -o Output/symtab.llvm.o

Note how it's missing the warning flag from the CMakeLists.txt file.

So I have no idea what's going on (I can't even really run the tests; I'm on Windows and test-suite requires perl to run).

In D122983#3463561, @MaskRay wrote:

Adding a global -Wno-implicit-function-declaration looks good. @Meinersbur

+1 to this idea, but I'm worried it won't fix that particular bot.

You probably need to fix Makefiles as well. I believe there were similar mysterious errors with -ffp-contract changes. Try to look at history of llvm test-suite.

cc @fhahn if he can remember more

[aaron.ballman]

In D122983#3464331, @xbolva00 wrote:

You probably need to fix Makefiles as well. I believe there were similar mysterious errors with -ffp-contract changes. Try to look at history of llvm test-suite.

cc @fhahn if he can remember more

Good call on looking back through the commit history. It took a while to find anyone mucking about with warning flags, but I noticed: https://github.com/llvm/llvm-test-suite/commit/24550c3385e8e3703ed364e1ce20b06de97bbeee so it looks like CFLAGS is used. I'll push a speculative fix that tries that.

[xbolva00]

https://github.com/llvm/llvm-test-suite/commit/79f2b03c5111392d647fa542e120f70c8f39a991

[aaron.ballman]

In D122983#3464342, @xbolva00 wrote:

https://github.com/llvm/llvm-test-suite/commit/79f2b03c5111392d647fa542e120f70c8f39a991

Awesome, parallel build systems to boot. I'll make that change at the same time as setting CFLAGS and hopefully the combination of changes will appease the bots.

[xbolva00]

In D122983#3464347, @aaron.ballman wrote:

In D122983#3464342, @xbolva00 wrote:

https://github.com/llvm/llvm-test-suite/commit/79f2b03c5111392d647fa542e120f70c8f39a991

Awesome, parallel build systems to boot. I'll make that change at the same time as setting CFLAGS and hopefully the combination of changes will appease the bots.

Anyway true pain to support both. Maybe time to open discussion to remove Makefile support as it is a maintenence hell.

[aaron.ballman]

In D122983#3464347, @aaron.ballman wrote:

In D122983#3464342, @xbolva00 wrote:

https://github.com/llvm/llvm-test-suite/commit/79f2b03c5111392d647fa542e120f70c8f39a991

Awesome, parallel build systems to boot. I'll make that change at the same time as setting CFLAGS and hopefully the combination of changes will appease the bots.

Speculative fixes landed in 33a2cab37889f9599d71d27d48895d0992fe9735

[fhahn]

In D122983#3464331, @xbolva00 wrote:

In D122983#3464315, @aaron.ballman wrote:

In D122983#3463561, @MaskRay wrote:

Adding a global -Wno-implicit-function-declaration looks good. @Meinersbur

+1 to this idea, but I'm worried it won't fix that particular bot.

You probably need to fix Makefiles as well. I believe there were similar mysterious errors with -ffp-contract changes. Try to look at history of llvm test-suite.

cc @fhahn if he can remember more

Depending on how many tests need updating in llvm-test-suite, I think it would be better to only disable the warning per-test, after verifying the warning works as expected.

[Jake-Egan]

Hi, we also have a failure on AIX with test-suite call to undeclared library function '%0' with type %1; ISO C99 and later https://lab.llvm.org/buildbot/#/builders/214/builds/825/steps/9/logs/stdio

[aaron.ballman]

In D122983#3464724, @Jake-Egan wrote:

Hi, we also have a failure on AIX with test-suite call to undeclared library function '%0' with type %1; ISO C99 and later https://lab.llvm.org/buildbot/#/builders/214/builds/825/steps/9/logs/stdio

Thanks for letting me know, this should now be fixed.

[daltenty]

In D122983#3464759, @aaron.ballman wrote:

In D122983#3464724, @Jake-Egan wrote:

Hi, we also have a failure on AIX with test-suite call to undeclared library function '%0' with type %1; ISO C99 and later https://lab.llvm.org/buildbot/#/builders/214/builds/825/steps/9/logs/stdio

Thanks for letting me know, this should now be fixed.

Looks like there are more test-suite case breaking beyond the one you fixed:

https://lab.llvm.org/buildbot/#/builders/214/builds/842

I suspect there will be more with the scope of this change unfortunately.

[aaron.ballman]

In D122983#3465052, @daltenty wrote:

In D122983#3464759, @aaron.ballman wrote:

In D122983#3464724, @Jake-Egan wrote:

Hi, we also have a failure on AIX with test-suite call to undeclared library function '%0' with type %1; ISO C99 and later https://lab.llvm.org/buildbot/#/builders/214/builds/825/steps/9/logs/stdio

Thanks for letting me know, this should now be fixed.

Looks like there are more test-suite case breaking beyond the one you fixed:

https://lab.llvm.org/buildbot/#/builders/214/builds/842

I suspect there will be more with the scope of this change unfortunately.

I agree there will be more -- it's incredibly frustrating that the AIX bot runs the test-suite but stops testing after the first failure. It makes it really hard to maintain. :-(

I've been keeping my eye on that bot and have been fixing up the tests as they get noticed, but if someone who has access to AIX wanted to tell me all of the test failures so I could hit them all at once (or disable the warnings in CMakeLists.txt and Makefile), that'd be greatly appreciated!

[nemanjai]

@daltenty Can you please run this with the same config as the bot on one of our AIX machines but be sure to pass -i if it's make or -k 0 if it's ninja?

[aaron.ballman]

In D122983#3465099, @nemanjai wrote:

@daltenty Can you please run this with the same config as the bot on one of our AIX machines but be sure to pass -i if it's make or -k 0 if it's ninja?

That would be *awesome*! I have to take off for a few hours, but I'm happy to fix all of the issues you spot in one go when I get back in a bit. Thank you!

[nemanjai]

In D122983#3465122, @aaron.ballman wrote:

In D122983#3465099, @nemanjai wrote:

@daltenty Can you please run this with the same config as the bot on one of our AIX machines but be sure to pass -i if it's make or -k 0 if it's ninja?

That would be *awesome*! I have to take off for a few hours, but I'm happy to fix all of the issues you spot in one go when I get back in a bit. Thank you!

Actually, I had a build ready on AIX and ran it. Looks like just 2 more:

test-suite :: MultiSource/Benchmarks/MiBench/network-patricia/network-patricia.test
test-suite :: MultiSource/Benchmarks/nbench/nbench.test

[Jake-Egan]

@aaron.ballman Here are all the errors. Thanks!

MultiSource/Applications/d/write_ctables.c:600:6: error: call to undeclared library function 'strncasecmp' with type 'int (const char *, const char *, unsigned long)'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/Prolangs-C/bison/conflicts.c:266:7: error: call to undeclared function 'bcopy'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/Prolangs-C/bison/symtab.c:65:3: error: call to undeclared library function 'strcpy' with type 'char *(char *, const char *)'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/Prolangs-C/bison/symtab.c:91:11: error: call to undeclared library function 'strcmp' with type 'int (const char *, const char *)'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/Ptrdist/anagram/anagram.c:317:5: error: call to undeclared library function 'bzero' with type 'void (void *, unsigned long)'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/Ptrdist/yacr2/maze.c:316:5: error: call to undeclared library function 'bzero' with type 'void (void *, unsigned long)'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/nbench/nbench1.c:2920:1: error: call to undeclared library function 'bzero' with type 'void (void *, unsigned long)'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/MiBench/network-patricia/patricia.c:146:5: error: call to undeclared function 'bcopy'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/MiBench/network-patricia/patricia.c:150:5: error: call to undeclared function 'bcopy'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/MiBench/network-patricia/patricia.c:156:4: error: call to undeclared function 'bcopy'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/MiBench/network-patricia/patricia.c:297:4: error: call to undeclared function 'bcopy'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
MultiSource/Benchmarks/MiBench/network-patricia/patricia_test.c:100:2: error: call to undeclared library function 'bzero' with type 'void (void *, unsigned long)'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
SingleSource/UnitTests/2005-05-11-Popcount-ffs-fls.c:110:69: error: call to undeclared function 'ffsl'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]

[aaron.ballman]

Thank you @nemanjai and @Jake-Egan for the help, I really appreciate it! I've disabled the diagnostics from another test (it looks like I caught the other ones earlier today), so hopefully the bot goes back to green shortly. If not, I'll keep on it. :-)

[aaron.ballman]

clang/lib/Sema/SemaDecl.cpp

15319–15321

I've fixed that up in a9d68a5524dea113cace5983697786599cbdce9a, thanks for the suggestion!

[manojgupta]

We are finding a lot of failures in our ToT builds with this change. here is an example for a configure script:

$ cat tent.c
int main ()
{
tgetent(0,0);
return 0;
}
$ bin/clang -c tent.c -Wno-error
tent.c:3:2: error: call to undeclared function 'tgetent'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
tgetent(0,0);
^
1 error generated.

It feels very surprising that Wno-error does not suppress this warning. Is that expected?

[xbolva00]

In D122983#3484064, @manojgupta wrote:

We are finding a lot of failures in our ToT builds with this change. here is an example for a configure script:

$ cat tent.c
int main ()
{
tgetent(0,0);
return 0;
}
$ bin/clang -c tent.c -Wno-error
tent.c:3:2: error: call to undeclared function 'tgetent'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
tgetent(0,0);
^
1 error generated.

It feels very surprising that Wno-error does not suppress this warning. Is that expected?

Yéah, this is unfortunate if -Wno-error does nothing :/ other options from release notes work for you?

[aaron.ballman]

In D122983#3484064, @manojgupta wrote:

We are finding a lot of failures in our ToT builds with this change. here is an example for a configure script:

$ cat tent.c
int main ()
{
tgetent(0,0);
return 0;
}
$ bin/clang -c tent.c -Wno-error
tent.c:3:2: error: call to undeclared function 'tgetent'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
tgetent(0,0);
^
1 error generated.

It feels very surprising that Wno-error does not suppress this warning. Is that expected?

Yes and no. Warnings which default to an error have very surprising behavior (at least to me) in terms of -Werror and -w. Specifying -Wno-error doesn't downgrade these diagnostics into a warning (https://godbolt.org/z/c43zTqTj1) and specifying -w doesn't disable the warning (https://godbolt.org/z/Y8YYYecTd). As you can see, that behavior is not specific to this patch but is just a general behavior with these kinds of diagnostics. I also find the behavior surprising (in both cases), but the last time I asked around about it, it sounded like this behavior was intentional for some reasons. More exploration is needed to know whether we should make any changes there.

In the meantime, -Wno-error=implicit-function-declaration should downgrade the error to a warning for you (but as you noticed, you can't later re-upgrade it into an error; so yes, these things behave a bit strangely IMO).

[manojgupta]

Hmm, the commit message says that Wno-error should work but this is not really the case :(.

(they can disable the warning or use -Wno-error to downgrade the
error

[rjmccall]

IIRC, the reason it works it that way is that "warnings which default to an error" are really "errors which you can explicitly downgrade to a warning". Maybe those ought to be different categories, or maybe we ought to just be telling people to downgrade this specific diagnostic instead of generally using -Wno-error.

[aaron.ballman]

In D122983#3484189, @rjmccall wrote:

IIRC, the reason it works it that way is that "warnings which default to an error" are really "errors which you can explicitly downgrade to a warning". Maybe those ought to be different categories, or maybe we ought to just be telling people to downgrade this specific diagnostic instead of generally using -Wno-error.

Thanks for the background John! That seems defensible, but I think we're still in a confused state that we should do something about. A warning which default to an error is not sufficiently warning-like to be disabled via -w but isn't sufficiently error-like to be disabled via -Wno-error. IMO, one or the other of those options should work with a diagnostic like this.

However, for the time being, the advice is to use -Wno-error=implicit-function-declaration, which is what's documented in the release note. The commit message had a shorthand suggesting -Wno-error, but my intent was to suggest people narrowly disable the error instead of broadly disable them all.

clang-tools-extra/clangd/IncludeFixer.cpp

std::vector<Fix> IncludeFixer::fix(DiagnosticsEngine::Level DiagLevel,
   case diag::err_undeclared_use_suggest:
   case diag::err_undeclared_var_use:
   case diag::err_undeclared_var_use_suggest:
   case diag::err_no_member: // Could be no member in namespace.
   case diag::err_no_member_suggest:
   case diag::err_no_member_template:
   case diag::err_no_member_template_suggest:
   case diag::warn_implicit_function_decl:
-  case diag::ext_implicit_function_decl:
+  case diag::ext_implicit_function_decl_c99:
+  case diag::ext_implicit_function_decl_c11:
   case diag::err_opencl_implicit_function_decl:
     dlog("Unresolved name at {0}, last typo was {1}",
          Info.getLocation().printToString(Info.getSourceManager()),
          LastUnresolvedName
              ? LastUnresolvedName->Loc.printToString(Info.getSourceManager())
              : "none");
     if (LastUnresolvedName) {
       // Try to fix unresolved name caused by missing declaration.

clang-tools-extra/clangd/ParsedAST.cpp

for (const auto &Check : CTChecks) {
       Check->registerMatchers(&CTFinder);
     }
 
     // Clang only corrects typos for use of undeclared functions in C if that
     // use is an error. Include fixer relies on typo correction, so pretend
     // this is an error. (The actual typo correction is nice too).
     // We restore the original severity in the level adjuster.
     // FIXME: It would be better to have a real API for this, but what?
-    for (auto ID : {diag::ext_implicit_function_decl,
+    for (auto ID : {diag::ext_implicit_function_decl_c99,
+                    diag::ext_implicit_function_decl_c11,
                     diag::warn_implicit_function_decl}) {
       OverriddenSeverity.try_emplace(
           ID, Clang->getDiagnostics().getDiagnosticLevel(ID, SourceLocation()));
       Clang->getDiagnostics().setSeverity(ID, diag::Severity::Error,
                                           SourceLocation());
     }
 
     const Config &Cfg = Config::current();

clang-tools-extra/test/clang-tidy/checkers/bugprone-easily-swappable-parameters-implicits.c

 // RUN: %check_clang_tidy %s bugprone-easily-swappable-parameters %t \
 // RUN:   -config='{CheckOptions: [ \
 // RUN:     {key: bugprone-easily-swappable-parameters.MinimumLength, value: 2}, \
 // RUN:     {key: bugprone-easily-swappable-parameters.IgnoredParameterNames, value: ""}, \
 // RUN:     {key: bugprone-easily-swappable-parameters.IgnoredParameterTypeSuffixes, value: ""}, \
 // RUN:     {key: bugprone-easily-swappable-parameters.QualifiersMix, value: 0}, \
 // RUN:     {key: bugprone-easily-swappable-parameters.ModelImplicitConversions, value: 1}, \
 // RUN:     {key: bugprone-easily-swappable-parameters.SuppressParametersUsedTogether, value: 0}, \
 // RUN:     {key: bugprone-easily-swappable-parameters.NamePrefixSuffixSilenceDissimilarityTreshold, value: 0} \
-// RUN:  ]}' --
+// RUN:  ]}' -- -std=c89
 
 void implicitDoesntBreakOtherStuff(int A, int B) {}
 // CHECK-MESSAGES: :[[@LINE-1]]:36: warning: 2 adjacent parameters of 'implicitDoesntBreakOtherStuff' of similar type ('int') are easily swapped by mistake [bugprone-easily-swappable-parameters]
 // CHECK-MESSAGES: :[[@LINE-2]]:40: note: the first parameter in the range is 'A'
 // CHECK-MESSAGES: :[[@LINE-3]]:47: note: the last parameter in the range is 'B'
 
 void arrayAndPtr1(int *IP, int IA[]) { arrayAndPtr1(IA, IP); }
 // CHECK-MESSAGES: :[[@LINE-1]]:19: warning: 2 adjacent parameters of 'arrayAndPtr1' of similar type ('int *')

 // CHECK-MESSAGES: :[[@LINE-4]]:35: note: 'double' and 'enum Unscoped' may be implicitly converted{{$}}
 
 void numericConversion8(double D, enum UnscopedFixed UF) { numericConversion8(UF, D); }
 // CHECK-MESSAGES: :[[@LINE-1]]:25: warning: 2 adjacent parameters of 'numericConversion8' of convertible types
 // CHECK-MESSAGES: :[[@LINE-2]]:32: note: the first parameter in the range is 'D'
 // CHECK-MESSAGES: :[[@LINE-3]]:54: note: the last parameter in the range is 'UF'
 // CHECK-MESSAGES: :[[@LINE-4]]:35: note: 'double' and 'enum UnscopedFixed' may be implicitly converted{{$}}
 
 void pointeeConverison(int *IP, double *DP) { pointeeConversion(DP, IP); }
 // NO-WARN: Even though this is possible in C, a swap is diagnosed by the compiler.

clang/docs/ReleaseNotes.rst

- ``-Wliteral-range`` will warn on floating-point equality comparisons with
   ``(float) f == 0.1`` is always false.
 - ``-Winline-namespace-reopened-noninline`` now takes into account that the
   ``inline`` keyword must appear on the original but not necessarily all
   extension definitions of an inline namespace and therefore points its note
   at the original definition. This fixes `Issue 50794 (PR51452)
   <https://github.com/llvm/llvm-project/issues/50794>`_.
 - ``-Wunused-but-set-variable`` now also warns if the variable is only used
   by unary operators.
+- The ``-Wimplicit-function-declaration`` warning diagnostic now defaults to
+  emitting an error (which can be downgraded to a warning with
+  ``-Wno-error=implicit-function-declaration`` or disabled entirely with
+  ``-Wno-implicit-function-declaration``) in C11 and
+  C17 mode. This is because the feature was removed in C99 and cannot be
+  supported in C2x.
 
 Non-comprehensive list of changes in this release
 -------------------------------------------------
 - Improve __builtin_dump_struct:
   - Support bitfields in struct and union.
   - Improve the dump format, dump both bitwidth(if its a bitfield) and field value.
   - Remove anonymous tag locations.
   - Beautify dump format, add indent for nested struct and struct members.

 -------------------
 
 - Implemented `WG14 N2674 The noreturn attribute <http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2764.pdf>`_.
 - Implemented `WG14 N2935 Make false and true first-class language features <http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2935.pdf>`_.
 - Implemented `WG14 N2763 Adding a fundamental type for N-bit integers <http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2763.pdf>`_.
 - Implemented `WG14 N2775 Literal suffixes for bit-precise integers <http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2775.pdf>`_.
 - Implemented the `*_WIDTH` macros to complete support for
   `WG14 N2412 Two's complement sign representation for C2x <https://www9.open-std.org/jtc1/sc22/wg14/www/docs/n2412.pdf>`_.
+- Removed support for implicit function declarations. This was a C89 feature
+  that was removed in C99, but cannot be supported in C2x because it requires
+  support for functions without prototypes, which no longer exist in C2x.
 
 C++ Language Changes in Clang
 -----------------------------
 
 - ...
 
 C++20 Feature Support
 ^^^^^^^^^^^^^^^^^^^^^

clang/include/clang/Basic/DiagnosticSemaKinds.td

def warn_return_value_udt: Warning<
   "%0 has C-linkage specified, but returns user-defined type %1 which is "
   "incompatible with C">, InGroup<ReturnTypeCLinkage>;
 def warn_return_value_udt_incomplete: Warning<
   "%0 has C-linkage specified, but returns incomplete type %1 which could be "
   "incompatible with C">, InGroup<ReturnTypeCLinkage>;
 def warn_implicit_function_decl : Warning<
   "implicit declaration of function %0">,
   InGroup<ImplicitFunctionDeclare>, DefaultIgnore;
-def ext_implicit_function_decl : ExtWarn<
+def ext_implicit_function_decl_c99 : ExtWarn<
   "implicit declaration of function %0 is invalid in C99">,
   InGroup<ImplicitFunctionDeclare>;
+def ext_implicit_function_decl_c11 : ExtWarn<
+  "implicit declaration of function %0 is invalid in C99 and later and "
+  "unsupported in C2x">,
+  InGroup<ImplicitFunctionDeclare>, DefaultError;
 def note_function_suggestion : Note<"did you mean %0?">;
 
 def err_ellipsis_first_param : Error<
   "ISO C requires a named parameter before '...'">;
 def err_declarator_need_ident : Error<"declarator requires an identifier">;
 def err_language_linkage_spec_unknown : Error<"unknown linkage language">;
 def err_language_linkage_spec_not_ascii : Error<
   "string literal in language linkage specifier cannot have an "
   "encoding-prefix">;
def warn_unreachable_loop_increment : Warning<
   InGroup<UnreachableCodeLoopIncrement>, DefaultIgnore;
 def warn_unreachable_fallthrough_attr : Warning<
   "fallthrough annotation in unreachable code">,
   InGroup<UnreachableCodeFallthrough>, DefaultIgnore;
 def note_unreachable_silence : Note<
   "silence by adding parentheses to mark code as explicitly dead">;
 
 /// Built-in functions.
 def ext_implicit_lib_function_decl : ExtWarn<
   "implicitly declaring library function '%0' with type %1">,
   InGroup<ImplicitFunctionDeclare>;
 def note_include_header_or_declare : Note<
   "include the header <%0> or explicitly provide a declaration for '%1'">;
 def note_previous_builtin_declaration : Note<"%0 is a builtin with type %1">;
 def warn_implicit_decl_no_jmp_buf
     : Warning<"declaration of built-in function '%0' requires the declaration"
     " of the 'jmp_buf' type, commonly provided in the header <setjmp.h>.">,
       InGroup<DiagGroup<"incomplete-setjmp-declaration">>;
 def warn_implicit_decl_requires_sysheader : Warning<

clang/lib/Sema/SemaDecl.cpp

       //   declaration is visible for this identifier, the identifier is
       //   implicitly declared exactly as if, in the innermost block containing
       //   the function call, the declaration
       //
       //     extern int identifier ();
       //
       //   appeared.
       //
-      // We also allow this in C99 as an extension.
-      if (NamedDecl *D = ImplicitlyDefineFunction(NameLoc, *Name, S))
-        return NameClassification::NonType(D);
+      // We also allow this in C99 as an extension. However, this is not
+      // allowed in C2x as there are no functions without prototypes there.
+      if (!getLangOpts().C2x) {
+        if (NamedDecl *D = ImplicitlyDefineFunction(NameLoc, *Name, S))
+          return NameClassification::NonType(D);
+      }
     }
 
     if (getLangOpts().CPlusPlus20 && SS.isEmpty() && NextToken.is(tok::less)) {
       // In C++20 onwards, this could be an ADL-only call to a function
       // template, and we're required to assume that this is a template name.
       //
       // FIXME: Find a way to still do typo correction in this case.
       TemplateName Template =

     if (Method->isStatic())
       checkThisInStaticMemberFunctionAttributes(Method);
 }
 
 /// ImplicitlyDefineFunction - An undeclared identifier was used in a function
 /// call, forming a call to an implicitly defined function (per C99 6.5.1p2).
 NamedDecl *Sema::ImplicitlyDefineFunction(SourceLocation Loc,
                                           IdentifierInfo &II, Scope *S) {
+  // It is not valid to implicitly define a function in C2x.
+  assert(!LangOpts.C2x && "Cannot implicitly define a function in C2x");
+
   // Find the scope in which the identifier is injected and the corresponding
   // DeclContext.
   // FIXME: C89 does not say what happens if there is no enclosing block scope.
   // In that case, we inject the declaration into the translation unit scope
   // instead.
   Scope *BlockScope = S;
   while (!BlockScope->isCompoundStmtScope() && BlockScope->getParent())
     BlockScope = BlockScope->getParent();

 
   // Extension in C99.  Legal in C90, but warn about it.
   unsigned diag_id;
   if (II.getName().startswith("__builtin_"))
     diag_id = diag::warn_builtin_unknown;
   // OpenCL v2.0 s6.9.u - Implicit function declaration is not supported.
   else if (getLangOpts().OpenCL)
     diag_id = diag::err_opencl_implicit_function_decl;
+  else if (getLangOpts().C11)
+    diag_id = diag::ext_implicit_function_decl_c11;
   else if (getLangOpts().C99)
-    diag_id = diag::ext_implicit_function_decl;
+    diag_id = diag::ext_implicit_function_decl_c99;
   else
     diag_id = diag::warn_implicit_function_decl;
 
   TypoCorrection Corrected;
   // Because typo correction is expensive, only do it if the implicit
   // function declaration is going to be treated as an error.
   //
   // Perform the corection before issuing the main diagnostic, as some consumers

     unsigned FormatIdx;
     bool HasVAListArg;
     if (Context.BuiltinInfo.isPrintfLike(BuiltinID, FormatIdx, HasVAListArg)) {
       if (!FD->hasAttr<FormatAttr>()) {
         const char *fmt = "printf";
         unsigned int NumParams = FD->getNumParams();
         if (FormatIdx < NumParams && // NumParams may be 0 (e.g. vfprintf)
             FD->getParamDecl(FormatIdx)->getType()->isObjCObjectPointerType())
           fmt = "NSString";
         FD->addAttr(FormatAttr::CreateImplicit(Context,
                                                &Context.Idents.get(fmt),
                                                FormatIdx+1,
                                                HasVAListArg ? 0 : FormatIdx+2,
                                                FD->getLocation()));
       }
     }
     if (Context.BuiltinInfo.isScanfLike(BuiltinID, FormatIdx,

       case Builtin::BIfma:
       case Builtin::BIfmaf:
       case Builtin::BIfmal:
         FD->addAttr(ConstAttr::CreateImplicit(Context, FD->getLocation()));
         break;
       default:
         break;
       }
     }
 
     if (Context.BuiltinInfo.isReturnsTwice(BuiltinID) &&
         !FD->hasAttr<ReturnsTwiceAttr>())
       FD->addAttr(ReturnsTwiceAttr::CreateImplicit(Context,
                                          FD->getLocation()));
     if (Context.BuiltinInfo.isNoThrow(BuiltinID) && !FD->hasAttr<NoThrowAttr>())
       FD->addAttr(NoThrowAttr::CreateImplicit(Context, FD->getLocation()));
     if (Context.BuiltinInfo.isPure(BuiltinID) && !FD->hasAttr<PureAttr>())
       FD->addAttr(PureAttr::CreateImplicit(Context, FD->getLocation()));
     if (Context.BuiltinInfo.isConst(BuiltinID) && !FD->hasAttr<ConstAttr>())

clang/lib/Sema/SemaExpr.cpp

         return Ex;
     }
   }
 
   if (R.isAmbiguous())
     return ExprError();
 
   // This could be an implicitly declared function reference (legal in C90,
-  // extension in C99, forbidden in C++).
-  if (R.empty() && HasTrailingLParen && II && !getLangOpts().CPlusPlus) {
+  // extension in C99, forbidden in C++ and C2x).
+  if (R.empty() && HasTrailingLParen && II && !getLangOpts().CPlusPlus &&
+      !getLangOpts().C2x) {
     NamedDecl *D = ImplicitlyDefineFunction(NameLoc, *II, S);
     if (D) R.addDecl(D);
   }
 
   // Determine whether this name might be a candidate for
   // argument-dependent lookup.
   bool ADL = UseArgumentDependentLookup(SS, R, HasTrailingLParen);
 

clang/test/ARCMT/objcmt-arc-cf-annotations.m

static __inline__ __attribute__((always_inline)) CFRange CFRangeMake(CFIndex loc, CFIndex len) {
     return range;
 }
 typedef const void * CFTypeRef;
 typedef const struct __CFString * CFStringRef;
 typedef const struct __CFAllocator * CFAllocatorRef;
 extern const CFAllocatorRef kCFAllocatorDefault;
 extern CFTypeRef CFRetain(CFTypeRef cf);
 extern void CFRelease(CFTypeRef cf);
+extern CFTypeRef CFAutorelease(CFTypeRef cf);
 extern CFTypeRef CFMakeCollectable(CFTypeRef cf);
 typedef struct {
 }
 CFArrayCallBacks;
 extern const CFArrayCallBacks kCFTypeArrayCallBacks;
 typedef const struct __CFArray * CFArrayRef;
 typedef struct __CFArray * CFMutableArrayRef;
 extern CFMutableArrayRef CFArrayCreateMutable(CFAllocatorRef allocator, CFIndex capacity, const CFArrayCallBacks *callBacks);

clang/test/ARCMT/objcmt-arc-cf-annotations.m.result

 extern CFTypeRef CFRetain(CFTypeRef cf);
 
 CF_IMPLICIT_BRIDGING_ENABLED
 
 extern void CFRelease(CFTypeRef cf);
 
 CF_IMPLICIT_BRIDGING_DISABLED
 
+extern CFTypeRef CFAutorelease(CFTypeRef cf);
 extern CFTypeRef CFMakeCollectable(CFTypeRef cf);
 typedef struct {
 }
 CFArrayCallBacks;
 extern const CFArrayCallBacks kCFTypeArrayCallBacks;
 typedef const struct __CFArray * CFArrayRef;
 typedef struct __CFArray * CFMutableArrayRef;
 extern CFMutableArrayRef CFArrayCreateMutable(CFAllocatorRef allocator, CFIndex capacity, const CFArrayCallBacks *callBacks) CF_RETURNS_RETAINED;

clang/test/Analysis/OSAtomic_mac.c

 // RUN: %clang_analyze_cc1 -w -analyzer-checker=core,debug.ExprInspection \
 // RUN:                    -analyzer-output=text -verify %s
 
+extern void clang_analyzer_eval(int);
+
 int OSAtomicCompareAndSwapPtrBarrier(void *, void *, void **);
 int OSAtomicCompareAndSwapPtrBarrier(void *, void *, void **) {
   // There is some body in the actual header,
   // but we should trust our BodyFarm instead.
 }
 
 int *invalidSLocOnRedecl(void) {
   // Was crashing when trying to throw a report about returning an uninitialized

clang/test/Analysis/ObjCProperties.m

 // RUN: %clang_analyze_cc1 -w %s -verify \
 // RUN:     -analyzer-checker=core,alpha.core,debug.ExprInspection
 
 #ifdef HEADER // A clever trick to avoid splitting up the test.
+extern void clang_analyzer_eval(int);
 
 @interface NSObject
 @end
 
 @interface HeaderClass : NSObject
 @property NSObject *prop;
 @end
 

clang/test/Analysis/PR49642.c

-// RUN: %clang_analyze_cc1 -w -verify %s \
+// RUN: %clang_analyze_cc1 -std=c99 -w -verify %s \
 // RUN:   -analyzer-checker=core \
 // RUN:   -analyzer-checker=apiModeling.StdCLibraryFunctions
 
 // expected-no-diagnostics
 
 typedef ssize_t;
 b;
 

clang/test/Analysis/diagnostics/no-store-func-path-notes.c

-// RUN: %clang_analyze_cc1 -w -x c -analyzer-checker=core -analyzer-output=text\
+// RUN: %clang_analyze_cc1 -w -std=c99 -analyzer-checker=core -analyzer-output=text\
 // RUN:     -verify %s
 
 typedef __typeof(sizeof(int)) size_t;
 void *memset(void *__s, int __c, size_t __n);
 
 int initializer1(int *p, int x) {
   if (x) { // expected-note{{'x' is 0}}
            // expected-note@-1{{Taking false branch}}

clang/test/Analysis/misc-ps-region-store.m

-// RUN: %clang_analyze_cc1 -triple i386-apple-darwin9 -analyzer-checker=core,alpha.core.CastToStruct,alpha.security.ReturnPtrRange,alpha.security.ArrayBound -analyzer-store=region -verify -fblocks -analyzer-opt-analyze-nested-blocks -Wno-objc-root-class -Wno-strict-prototypes %s
-// RUN: %clang_analyze_cc1 -triple x86_64-apple-darwin9 -DTEST_64 -analyzer-checker=core,alpha.core.CastToStruct,alpha.security.ReturnPtrRange,alpha.security.ArrayBound -analyzer-store=region -verify -fblocks   -analyzer-opt-analyze-nested-blocks -Wno-objc-root-class -Wno-strict-prototypes %s
+// RUN: %clang_analyze_cc1 -std=c99 -triple i386-apple-darwin9 -analyzer-checker=core,alpha.core.CastToStruct,alpha.security.ReturnPtrRange,alpha.security.ArrayBound -analyzer-store=region -verify -fblocks -analyzer-opt-analyze-nested-blocks -Wno-objc-root-class -Wno-strict-prototypes %s
+// RUN: %clang_analyze_cc1 -std=c99 -triple x86_64-apple-darwin9 -DTEST_64 -analyzer-checker=core,alpha.core.CastToStruct,alpha.security.ReturnPtrRange,alpha.security.ArrayBound -analyzer-store=region -verify -fblocks   -analyzer-opt-analyze-nested-blocks -Wno-objc-root-class -Wno-strict-prototypes %s
 
 typedef long unsigned int size_t;
 void *memcpy(void *, const void *, size_t);
 void *alloca(size_t);
 
 typedef struct objc_selector *SEL;
 typedef signed char BOOL;
 typedef int NSInteger;

clang/test/Analysis/novoidtypecrash.c

-// RUN: %clang_analyze_cc1 -analyzer-checker=core %s
+// RUN: %clang_analyze_cc1 -std=c89 -analyzer-checker=core %s
 x;
 y(void **z) { // no-crash
   *z = x;
   int *w;
   y(&w);
   *w;
 }
 

clang/test/Analysis/plist-macros-with-expansion.c

-// RUN: %clang_analyze_cc1 -analyzer-checker=core %s  \
+// RUN: %clang_analyze_cc1 -std=c99 -analyzer-checker=core %s  \
 // RUN:   -analyzer-output=plist -o %t.plist \
 // RUN:   -analyzer-config expand-macros=true -verify
 //
 // RUN: FileCheck --input-file=%t.plist %s
 
 #define STRANGE_FN(x) STRANGE_FN(x, 0)
 void test_strange_macro_expansion(void) {
   char *path;

clang/test/CodeGen/2002-07-14-MiscTests3.c

-// RUN: %clang_cc1 -w -emit-llvm %s  -o /dev/null
+// RUN: %clang_cc1 -std=c99 -w -emit-llvm %s  -o /dev/null
 
 void *malloc(unsigned);
 int puts(const char *s);
 
 struct FunStructTest {
   int Test1;
   char *Pointer;
   int Array[12];

clang/test/CodeGen/2002-07-31-SubregFailure.c

-// RUN: %clang_cc1 -emit-llvm %s  -o /dev/null
+// RUN: %clang_cc1 -std=c99 -emit-llvm %s  -o /dev/null
 
 
 typedef union {
    long (*ap)[4];
 } ptrs;
 
 void DoAssignIteration(void) {
   ptrs abase;
   abase.ap+=27;
   Assignment(*abase.ap);
 }
 
 

clang/test/CodeGen/2003-08-18-SigSetJmp.c

-// RUN: %clang_cc1 -triple x86_64-apple-darwin -emit-llvm %s  -o /dev/null
+// RUN: %clang_cc1 -triple x86_64-apple-darwin -std=c99 -emit-llvm %s  -o /dev/null
 
 #define _JBLEN ((9 * 2) + 3 + 16)
 typedef int sigjmp_buf[_JBLEN + 1];
 int sigsetjmp(sigjmp_buf env, int savemask);
 sigjmp_buf B;
 int foo(void) {
   sigsetjmp(B, 1);
   bar();
 }

clang/test/CodeGen/2004-11-27-StaticFunctionRedeclare.c

-// RUN: %clang_cc1 -emit-llvm %s -o - | FileCheck %s
+// RUN: %clang_cc1 -std=c89 -emit-llvm %s -o - | FileCheck %s
 
 // There should not be an unresolved reference to func here.  Believe it or not,
 // the "expected result" is a function named 'func' which is internal and
 // referenced by bar().
 
 // This is PR244
 
 // CHECK-LABEL: define {{.*}}void @bar(
 // CHECK: call {{.*}} @func
 // CHECK: define internal {{.*}}i32 @func(
 static int func();
 void bar(void) {
   int func(void);
   foo(func);
 }
 static int func(char** A, char ** B) {}

clang/test/CodeGen/2005-01-02-ConstantInits.c

 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --check-globals --global-value-regex "@.+"
-// RUN: %clang_cc1 -triple=x86_64-unknown-linux %s -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 -triple=x86_64-unknown-linux %s -std=c99 -emit-llvm -o - | FileCheck %s
 
 // This tests all kinds of hard cases with initializers and
 // array subscripts.  This corresponds to PR487.
 
 struct X { int a[2]; };
 
 //.
 // CHECK: @test.i23 = internal global i32 4, align 4

clang/test/CodeGen/2005-01-02-VAArgError-ICE.c

 // This file is erroneous, but should not cause the compiler to ICE.
 // PR481
-// RUN: %clang_cc1 %s -emit-llvm -o /dev/null
+// RUN: %clang_cc1 %s -std=c99 -emit-llvm -o /dev/null
 
 int flags(int a, int b, ...) {
         __builtin_va_list         args;
         __builtin_va_start(args,a);       // not the last named arg
         foo(args);
 }

clang/test/CodeGen/2006-01-13-StackSave.c

 // PR691
-// RUN: %clang_cc1 %s -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -std=c99 -emit-llvm -o - | FileCheck %s
 // CHECK: call i8* @llvm.stacksave()
 
 void test(int N) {
   int i;
   for (i = 0; i < N; ++i) {
     int VLA[i];
     external(VLA);
   }
 }

clang/test/CodeGen/2006-03-03-MissingInitializer.c

-// RUN: %clang_cc1 %s -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -emit-llvm -std=c89 -o - | FileCheck %s
 
 struct X { int *XX; int Y;};
 
 void foo(void) {
   // CHECK: @foo.nate = internal global i32 0
   static int nate = 0;
   struct X bob = { &nate, 14 };
   bar(&bob);
 }

clang/test/CodeGen/2008-05-12-TempUsedBeforeDef.c

-// RUN: %clang_cc1 -w -emit-llvm -o /dev/null %s
+// RUN: %clang_cc1 -std=c89 -w -emit-llvm -o /dev/null %s
 // PR2264.
 unsigned foo = 8L;
 unsigned bar = 0L;
 volatile unsigned char baz = 6L;
 int test(void) {
   char qux = 1L;
   for (; baz >= -29; baz--)
     bork(bar && foo, qux);
 }

clang/test/CodeGen/2008-07-30-redef-of-bitcasted-decl.c

-// RUN: %clang_cc1 -emit-llvm -o - %s
+// RUN: %clang_cc1 -std=c89 -emit-llvm -o - %s
 // <rdar://problem/6108358>
 
 /* For posterity, the issue here begins initial "char []" decl for
  * s. This is a tentative definition and so a global was being
  * emitted, however the mapping in GlobalDeclMap referred to a bitcast
  * of this global.
  *
  * The problem was that later when the correct definition for s is

clang/test/CodeGen/2008-08-19-cast-of-typedef.c

-// RUN: %clang_cc1 -emit-llvm -o %t %s
+// RUN: %clang_cc1 -std=c89 -emit-llvm -o %t %s
 
 typedef short T[4];
 struct s {
   T f0;
 };
 
 void foo(struct s *x) {
   bar((long) x->f0);
 }

clang/test/CodeGen/2008-10-13-FrontendCrash.c

-// RUN: %clang_cc1 %s -emit-llvm -o -
+// RUN: %clang_cc1 %s -std=c89 -emit-llvm -o -
 // PR2797
 
 unsigned int
 func_48 (signed char p_49)
 {
   signed char l_340;
   func_44 (1&((1 ^ 1 == (lshift_u_s (1)) != (l_340 < 1)) & 1L));
 }

clang/test/CodeGen/PowerPC/builtins-ppc-p8vector.c

 // REQUIRES: powerpc-registered-target
 // RUN: %clang_cc1 -target-feature +altivec -target-feature +power8-vector -triple powerpc64-unknown-unknown -emit-llvm %s -o - | FileCheck %s
 // RUN: %clang_cc1 -target-feature +altivec -target-feature +power8-vector -triple powerpc64le-unknown-unknown -emit-llvm %s -o - | FileCheck %s -check-prefix=CHECK-LE
-// RUN: not %clang_cc1 -target-feature +altivec -target-feature +vsx -triple powerpc64-unknown-unknown -emit-llvm %s -o - 2>&1 | FileCheck %s -check-prefix=CHECK-PPC
+// RUN: not %clang_cc1 -std=c99 -target-feature +altivec -target-feature +vsx -triple powerpc64-unknown-unknown -emit-llvm %s -o - 2>&1 | FileCheck %s -check-prefix=CHECK-PPC
 // Added -target-feature +vsx above to avoid errors about "vector double" and to
 // generate the correct errors for functions that are only overloaded with VSX
 // (vec_cmpge, vec_cmple). Without this option, there is only one overload so
 // it is selected.
 #include <altivec.h>
 
 void dummy() { }
 signed int si;

clang/test/CodeGen/X86/bmi2-builtins.c

-// RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +bmi2 -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +bmi2 -emit-llvm -std=c99 -o - | FileCheck %s
 // RUN: %clang_cc1 -ffreestanding %s -triple=i386-apple-darwin -target-feature +bmi2 -emit-llvm -o - | FileCheck %s --check-prefix=B32
 
 
 #include <immintrin.h>
 
 unsigned int test_bzhi_u32(unsigned int __X, unsigned int __Y) {
   // CHECK: @llvm.x86.bmi.bzhi.32
   return _bzhi_u32(__X, __Y);

clang/test/CodeGen/aarch64-mops.c

 // RUN: %clang_cc1 -triple aarch64-arm-unknown-eabi -target-feature +mops -target-feature +mte -w -S -emit-llvm -o - %s  | FileCheck --check-prefix=CHECK-MOPS   %s
-// RUN: %clang_cc1 -triple aarch64-arm-unknown-eabi -target-feature +mops                      -w -S -emit-llvm -o - %s  | FileCheck --check-prefix=CHECK-NOMOPS %s
-// RUN: %clang_cc1 -triple aarch64-arm-unknown-eabi                       -target-feature +mte -w -S -emit-llvm -o - %s  | FileCheck --check-prefix=CHECK-NOMOPS %s
-// RUN: %clang_cc1 -triple aarch64-arm-unknown-eabi                                            -w -S -emit-llvm -o - %s  | FileCheck --check-prefix=CHECK-NOMOPS %s
+// RUN: %clang_cc1 -triple aarch64-arm-unknown-eabi -target-feature +mops -std=c99             -w -S -emit-llvm -o - %s  | FileCheck --check-prefix=CHECK-NOMOPS %s
+// RUN: %clang_cc1 -triple aarch64-arm-unknown-eabi -std=c99              -target-feature +mte -w -S -emit-llvm -o - %s  | FileCheck --check-prefix=CHECK-NOMOPS %s
+// RUN: %clang_cc1 -triple aarch64-arm-unknown-eabi -std=c99                                   -w -S -emit-llvm -o - %s  | FileCheck --check-prefix=CHECK-NOMOPS %s
 
 #include <arm_acle.h>
 #include <stddef.h>
 
 // CHECK-LABEL:       @bzero_0(
 // CHECK-MOPS:        @llvm.aarch64.mops.memset.tag
 // CHECK-NOMOPS-NOT:  @llvm.aarch64.mops.memset.tag
 void *bzero_0(void *dst) {

clang/test/CodeGen/aarch64-neon-sm4-sm3.c

 // RUN: %clang_cc1 -triple aarch64-linux-gnu -target-feature +neon \
 // RUN:  -target-feature +sm4 -S -emit-llvm -o - %s \
 // RUN:  | FileCheck %s
 
-// RUN: not %clang_cc1 -triple aarch64-linux-gnu -target-feature +neon \
+// RUN: not %clang_cc1 -std=c99 -triple aarch64-linux-gnu -target-feature +neon \
 // RUN: -S -emit-llvm -o - %s 2>&1 | FileCheck --check-prefix=CHECK-NO-CRYPTO %s
 
 // REQUIRES: aarch64-registered-target || arm-registered-target
 
 #include <arm_neon.h>
 
 void test_vsm3partw1(uint32x4_t a, uint32x4_t b, uint32x4_t c) {
   // CHECK-LABEL: @test_vsm3partw1(

clang/test/CodeGen/arm_acle.c

 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
-// RUN: %clang_cc1 -ffreestanding -triple armv8a-none-eabi -target-feature +crc -target-feature +dsp -O0 -disable-O0-optnone -fexperimental-new-pass-manager -S -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s -check-prefixes=ARM,AArch32
-// RUN: %clang_cc1 -ffreestanding -triple aarch64-none-eabi -target-feature +neon -target-feature +crc -target-feature +crypto -O0 -disable-O0-optnone -fexperimental-new-pass-manager -S -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s -check-prefixes=ARM,AArch64
-// RUN: %clang_cc1 -ffreestanding -triple aarch64-none-eabi -target-feature +v8.3a -O0 -disable-O0-optnone -fexperimental-new-pass-manager -S -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s -check-prefixes=ARM,AArch64,AArch6483
-// RUN: %clang_cc1 -ffreestanding -triple aarch64-none-eabi -target-feature +v8.5a -target-feature +rand -O0 -disable-O0-optnone -fexperimental-new-pass-manager -S -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s -check-prefixes=ARM,AArch64,AArch6483,AArch6485
+// RUN: %clang_cc1 -std=c99 -ffreestanding -triple armv8a-none-eabi -target-feature +crc -target-feature +dsp -O0 -disable-O0-optnone -fexperimental-new-pass-manager -S -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s -check-prefixes=ARM,AArch32
+// RUN: %clang_cc1 -std=c99 -ffreestanding -triple aarch64-none-eabi -target-feature +neon -target-feature +crc -target-feature +crypto -O0 -disable-O0-optnone -fexperimental-new-pass-manager -S -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s -check-prefixes=ARM,AArch64
+// RUN: %clang_cc1 -std=c99 -ffreestanding -triple aarch64-none-eabi -target-feature +v8.3a -O0 -disable-O0-optnone -fexperimental-new-pass-manager -S -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s -check-prefixes=ARM,AArch64,AArch6483
+// RUN: %clang_cc1 -std=c99 -ffreestanding -triple aarch64-none-eabi -target-feature +v8.5a -target-feature +rand -O0 -disable-O0-optnone -fexperimental-new-pass-manager -S -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s -check-prefixes=ARM,AArch64,AArch6483,AArch6485
 
 #include <arm_acle.h>
 
 // REQUIRES: arm-registered-target,aarch64-registered-target
 
 /* 8 SYNCHRONIZATION, BARRIER AND HINT INTRINSICS */
 /* 8.3 Memory Barriers */
 

clang/test/CodeGen/attribute_constructor.c

 // RUN: %clang_cc1 %s -emit-llvm -o - | grep llvm.global_ctors
 
+extern int bar();
 void foo(void) __attribute__((constructor));
 void foo(void) {
   bar();
 }

clang/test/CodeGen/builtins-arm-microsoft.c

 // RUN: %clang_cc1 -triple thumbv7-windows -fms-compatibility -emit-llvm -o - %s \
 // RUN:     | FileCheck %s -check-prefix CHECK-MSVC
-// RUN: %clang_cc1 -triple armv7-eabi -emit-llvm %s -o - \
+// RUN: %clang_cc1 -std=c99 -triple armv7-eabi -emit-llvm %s -o - \
 // RUN:     | FileCheck %s -check-prefix CHECK-EABI
 // REQUIRES: arm-registered-target
 
 void test_yield_intrinsic() {
   __yield();
 }
 
 // CHECK-MSVC: call void @llvm.arm.hint(i32 1)

clang/test/CodeGen/builtins-arm-msvc-compat-only.c

 // RUN: %clang_cc1 -triple thumbv7-windows -fms-extensions -emit-llvm -o - %s \
 // RUN:     | FileCheck %s -check-prefix CHECK-MSVC
-// RUN: %clang_cc1 -triple armv7-eabi -emit-llvm %s -o /dev/null 2>&1 \
+// RUN: %clang_cc1 -std=c99 -triple armv7-eabi -emit-llvm %s -o /dev/null 2>&1 \
 // RUN:     | FileCheck %s -check-prefix CHECK-EABI
 // REQUIRES: arm-registered-target
 
 void emit() {
   __emit(0xdefe);
 }
 
 // CHECK-MSVC: call void asm sideeffect ".inst.n 0xDEFE", ""()
 // CHECK-EABI: warning: implicit declaration of function '__emit' is invalid in C99
 
 void emit_truncated() {
   __emit(0x11110000); // movs r0, r0
 }
 
 // CHECK-MSVC: call void asm sideeffect ".inst.n 0x0", ""()
 

clang/test/CodeGen/cast-emit.c

 // RUN: %clang_cc1 -emit-llvm %s -o - | FileCheck %s
 
+extern int f();
 typedef union {
   int    i;
   float  f;
 } MyUnion;
 void unionf(MyUnion a);
 void uniontest(float a) {
   f((MyUnion)1.0f);
 // CHECK: store float 1.000000e+00
 }
 

clang/test/CodeGen/debug-info-block-vars.c

-// RUN: %clang_cc1 -x c -fblocks -debug-info-kind=standalone -emit-llvm -O0 \
+// RUN: %clang_cc1 -std=c89 -fblocks -debug-info-kind=standalone -emit-llvm -O0 \
 // RUN:   -triple x86_64-apple-darwin -o - %s | FileCheck %s
-// RUN: %clang_cc1 -x c -fblocks -debug-info-kind=standalone -emit-llvm -O1 \
+// RUN: %clang_cc1 -std=c89 -fblocks -debug-info-kind=standalone -emit-llvm -O1 \
 // RUN:   -triple x86_64-apple-darwin -o - %s \
 // RUN:   | FileCheck --check-prefix=CHECK-OPT %s
 
 // CHECK: define internal void @__f_block_invoke(i8* noundef %.block_descriptor)
 // CHECK: %.block_descriptor.addr = alloca i8*, align 8
 // CHECK: %block.addr = alloca <{ i8*, i32, i32, i8*, %struct.__block_descriptor* }>*, align 8
 // CHECK: store i8* %.block_descriptor, i8** %.block_descriptor.addr, align 8
 // CHECK: call void @llvm.dbg.declare(metadata i8** %.block_descriptor.addr,

clang/test/CodeGen/debug-info-crash.c

 // REQUIRES: x86-registered-target
-// RUN: %clang_cc1 -triple i386-apple-darwin10 -fblocks -debug-info-kind=limited -S %s -o -
+// RUN: %clang_cc1 -std=c89 -triple i386-apple-darwin10 -fblocks -debug-info-kind=limited -S %s -o -
 
 // rdar://7590323
 typedef struct dispatch_queue_s *dispatch_queue_t;
 __attribute__((visibility("default")))
 extern struct dispatch_queue_s _dispatch_main_q;
 typedef struct dispatch_item_s *dispatch_item_t;
 typedef void (^dispatch_legacy_block_t)(dispatch_item_t);
 dispatch_item_t LEGACY_dispatch_call(dispatch_queue_t dq,

clang/test/CodeGen/decl.c

-// RUN: %clang_cc1 -w -fmerge-all-constants -emit-llvm < %s | FileCheck %s
+// RUN: %clang_cc1 -std=c89 -w -fmerge-all-constants -emit-llvm < %s | FileCheck %s
 
 // CHECK: @test1.x = internal constant [12 x i32] [i32 1
 // CHECK: @__const.test2.x = private unnamed_addr constant [13 x i32] [i32 1,
 // CHECK: @test5w = {{(dso_local )?}}global { i32, [4 x i8] } { i32 2, [4 x i8] undef }
 // CHECK: @test5y = {{(dso_local )?}}global { double } { double 7.300000e+0{{[0]*}}1 }
 
 // CHECK: @__const.test6.x = private unnamed_addr constant %struct.SelectDest { i8 1, i8 2, i32 3, i32 0 }
 

clang/test/CodeGen/init-with-member-expr.c

 // RUN: %clang_cc1 < %s -emit-llvm
 struct test {
   int a;
 };
 
 extern struct test t;
 
 int *b=&t.a;
 
 
 // PR2049
 typedef struct mark_header_tag {
  unsigned char mark[7];
 } mark_header_t;
+extern int foo();
 int is_rar_archive(int fd) {
         const mark_header_t rar_hdr[2] = {{0x52, 0x61, 0x72, 0x21, 0x1a, 0x07, 0x00}, {'U', 'n', 'i', 'q', 'u', 'E', '!'}};
         foo(rar_hdr);
 
         return 0;
 }
 

clang/test/CodeGen/misaligned-param.c

-// RUN: %clang_cc1 %s -triple i386-apple-darwin -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -std=c89 -triple i386-apple-darwin -emit-llvm -o - | FileCheck %s
 // Misaligned parameter must be memcpy'd to correctly aligned temporary.
 
 struct s { int x; long double y; };
 long double foo(struct s x, int i, struct s y) {
 // CHECK: foo
 // CHECK: %x = alloca %struct.s, align 16
 // CHECK: %y = alloca %struct.s, align 16
 // CHECK: memcpy
 // CHECK: memcpy
 // CHECK: bar
   return bar(&x, &y);
 }

clang/test/CodeGen/neon-crypto.c

 // RUN: %clang_cc1 -triple arm-none-linux-gnueabi -target-feature +neon \
 // RUN:  -target-feature +sha2 -target-feature +aes \
 // RUN:  -target-cpu cortex-a57 -emit-llvm -O1 -o - %s | FileCheck %s
 
 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
 // RUN:   -target-feature +sha2 -target-feature +aes \
 // RUN:   -emit-llvm -O1 -o - %s | FileCheck %s
-// RUN: not %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
+// RUN: not %clang_cc1 -std=c99 -triple arm64-none-linux-gnu -target-feature +neon \
 // RUN:   -S -O3 -o - %s 2>&1 | FileCheck --check-prefix=CHECK-NO-CRYPTO %s
 
 // REQUIRES: aarch64-registered-target || arm-registered-target
 
 #include <arm_neon.h>
 
 uint8x16_t test_vaeseq_u8(uint8x16_t data, uint8x16_t key) {
   // CHECK-LABEL: @test_vaeseq_u8

clang/test/CodeGen/struct-comma.c

 // RUN: %clang_cc1 %s -emit-llvm -o -
 
 struct S {int a, b;} x;
+extern int r(void);
 void a(struct S* b) {*b = (r(), x);}

clang/test/CodeGen/variable-array.c

-// RUN: %clang_cc1 -emit-llvm < %s | grep puts | count 4
+// RUN: %clang_cc1 -std=c99 -emit-llvm < %s | grep puts | count 4
 
 // PR3248
 int a(int x)
 {
   int (*y)[x];
   return sizeof(*(puts("asdf"),y));
 }
 

clang/test/Frontend/warning-mapping-2.c

 // Check that -w takes precedence over -pedantic-errors.
-// RUN: %clang_cc1 -verify -pedantic-errors -w %s
+// RUN: %clang_cc1 -verify -std=c89 -pedantic-errors -w %s
 
 // Expect *not* to see a diagnostic for "implicit declaration of function"
 // expected-no-diagnostics
 
 void f0(void) { f1(); }

clang/test/Headers/arm-cmse-header-ns.c

-// RUN:     %clang_cc1 -triple thumbv8m.base-eabi -fsyntax-only        %s 2>&1 | FileCheck --check-prefix=CHECK-c %s
+// RUN:     %clang_cc1 -triple thumbv8m.base-eabi -fsyntax-only -std=c99 %s 2>&1 | FileCheck --check-prefix=CHECK-c %s
 // RUN: not %clang_cc1 -triple thumbv8m.base-eabi -fsyntax-only -x c++ %s 2>&1 | FileCheck --check-prefix=CHECK-cpp %s
 
 #include <arm_cmse.h>
 
 typedef void (*callback_t)(void);
 
 void func(callback_t fptr, void *p)
 {

clang/test/Import/objc-arc/test-cleanup-object.m

 // UNSUPPORTED: -zos, -aix
 // RUN: clang-import-test -x objective-c -objc-arc -import %S/Inputs/cleanup-objects.m -dump-ast -expression %s | FileCheck %s
 
 // CHECK: FunctionDecl {{.*}} getObj '
 // CHECK: ExprWithCleanups
 // CHECK-NEXT: cleanup CompoundLiteralExpr
 
+extern int getObj();
 void test(int c, id a) {
   (void)getObj(c, a);
 }
 

clang/test/Modules/config_macros.m

 #undef WANT_FOO
 #define WANT_FOO 1
 @import config; // okay
 
 #define WANT_BAR 1 // expected-note{{macro was defined here}}
 @import config; // expected-warning{{definition of configuration macro 'WANT_BAR' has no effect on the import of 'config'; pass '-DWANT_BAR=...' on the command line to configure the module}}
 
 // RUN: rm -rf %t
-// RUN: %clang_cc1 -fmodules -fimplicit-module-maps -x objective-c -fmodules-cache-path=%t -DWANT_FOO=1 -emit-module -fmodule-name=config %S/Inputs/module.map
-// RUN: %clang_cc1 -fmodules -fimplicit-module-maps -fmodules-cache-path=%t -I %S/Inputs -DWANT_FOO=1 %s -verify
+// RUN: %clang_cc1 -std=c99 -fmodules -fimplicit-module-maps -x objective-c -fmodules-cache-path=%t -DWANT_FOO=1 -emit-module -fmodule-name=config %S/Inputs/module.map
+// RUN: %clang_cc1 -std=c99 -fmodules -fimplicit-module-maps -fmodules-cache-path=%t -I %S/Inputs -DWANT_FOO=1 %s -verify
 

clang/test/Modules/modulemap-locations.m

 // RUN: rm -rf %t 
-// RUN: %clang_cc1 -fmodules -fimplicit-module-maps -fmodules-cache-path=%t -I %S/Inputs/ModuleMapLocations/Module_ModuleMap -I %S/Inputs/ModuleMapLocations/Both -F %S/Inputs/ModuleMapLocations -I %S/Inputs/ModuleMapLocations -F %S/Inputs -x objective-c -fsyntax-only %s -verify -Wno-private-module
+// RUN: %clang_cc1 -std=c99 -fmodules -fimplicit-module-maps -fmodules-cache-path=%t -I %S/Inputs/ModuleMapLocations/Module_ModuleMap -I %S/Inputs/ModuleMapLocations/Both -F %S/Inputs/ModuleMapLocations -I %S/Inputs/ModuleMapLocations -F %S/Inputs -x objective-c -fsyntax-only %s -verify -Wno-private-module
 
 // regular
 @import module_modulemap;
 @import both;
 // framework
 @import Module_ModuleMap_F;
 @import Module_ModuleMap_F.Private;
 @import Both_F;
 @import Inferred;
 
 void test(void) {
   will_be_found1();
   wont_be_found1(); // expected-warning{{implicit declaration of function 'wont_be_found1' is invalid in C99}}
   will_be_found2();
   wont_be_found2(); // expected-warning{{implicit declaration of function 'wont_be_found2' is invalid in C99}}
 }

clang/test/OpenMP/declare_mapper_messages.c

 #pragma omp declare mapper(id: struct vec v) map(v.len) allocate(v)   // expected-error {{unexpected OpenMP clause 'allocate' in directive '#pragma omp declare mapper'}}
       struct vec vv, v1;
       struct vec arr[10];
       double d;
 #pragma omp target map(mapper)                                          // expected-error {{use of undeclared identifier 'mapper'}}
       {}
 #pragma omp target map(mapper:vv)                                       // expected-error {{expected '(' after 'mapper'}}
       {}
-#pragma omp target map(mapper( :vv)                                     // expected-error {{expected expression}} expected-error {{expected ')'}} expected-warning {{implicit declaration of function 'mapper' is invalid in C99}} expected-note {{to match this '('}}
+#pragma omp target map(mapper( :vv)                                     // expected-error {{expected expression}} expected-error {{expected ')'}} expected-error {{implicit declaration of function 'mapper' is invalid in C99}} expected-note {{to match this '('}}
       {}
-#pragma omp target map(mapper(aa :vv)                                   // expected-error {{use of undeclared identifier 'aa'}} expected-error {{expected ')'}} expected-warning {{implicit declaration of function 'mapper' is invalid in C99}} expected-note {{to match this '('}}
+#pragma omp target map(mapper(aa :vv)                                   // expected-error {{use of undeclared identifier 'aa'}} expected-error {{expected ')'}} expected-error {{implicit declaration of function 'mapper' is invalid in C99}} expected-note {{to match this '('}}
       {}
 #pragma omp target map(mapper(ab) :vv)                                  // expected-error {{missing map type}} expected-error {{cannot find a valid user-defined mapper for type 'struct vec' with name 'ab'}}
       {}
 #pragma omp target map(mapper(ab) :arr[0:2])                            // expected-error {{missing map type}} expected-error {{cannot find a valid user-defined mapper for type 'struct vec' with name 'ab'}}
       {}
 #pragma omp target map(mapper(aa) :vv)                                  // expected-error {{missing map type}}
       {}
 #pragma omp target map(mapper(aa) to:d)                                 // expected-error {{mapper type must be of struct, union or class type}}

clang/test/PCH/chain-macro-override.c

 // Test this without pch.
-// RUN: %clang_cc1 -include %S/Inputs/chain-macro-override1.h -include %S/Inputs/chain-macro-override2.h -fsyntax-only -verify -detailed-preprocessing-record %s
+// RUN: %clang_cc1 -std=c99 -include %S/Inputs/chain-macro-override1.h -include %S/Inputs/chain-macro-override2.h -fsyntax-only -verify -detailed-preprocessing-record %s
 
 // Test with pch.
-// RUN: %clang_cc1 -emit-pch -o %t1 %S/Inputs/chain-macro-override1.h -detailed-preprocessing-record 
-// RUN: %clang_cc1 -emit-pch -o %t2 %S/Inputs/chain-macro-override2.h -include-pch %t1 -detailed-preprocessing-record 
-// RUN: %clang_cc1 -include-pch %t2 -fsyntax-only -verify %s
+// RUN: %clang_cc1 -std=c99 -emit-pch -o %t1 %S/Inputs/chain-macro-override1.h -detailed-preprocessing-record 
+// RUN: %clang_cc1 -std=c99 -emit-pch -o %t2 %S/Inputs/chain-macro-override2.h -include-pch %t1 -detailed-preprocessing-record 
+// RUN: %clang_cc1 -std=c99 -include-pch %t2 -fsyntax-only -verify %s
 
 int foo(void) {
   f();
   g();
   h();
   h2(); // expected-warning{{implicit declaration of function 'h2' is invalid in C99}}
   h3();
   return x;
 }

clang/test/Rewriter/rewrite-foreach-2.m

 int INNERLOOP(void);
 void END_LOOP(void);
 @implementation MyList (BasicTest)
 - (void)compilerTestAgainst {
   id el;
         for (el in self) 
     { LOOP(); 
             for (id el1 in self) 
-         INNER_LOOP();
+         INNERLOOP();
 
       END_LOOP();
     }
 }
 @end
 

clang/test/Rewriter/rewrite-try-catch.m

-// RUN: %clang_cc1 -rewrite-objc -fobjc-runtime=macosx-fragile-10.5  %s -o -
+// RUN: %clang_cc1 -rewrite-objc -fobjc-runtime=macosx-fragile-10.5 -std=c99 %s -o -
 
 @interface Foo @end
 @interface GARF @end
 
 void foo(void) {
   @try  { TRY(); } 
   @catch (...) { SPLATCH(); @throw; }
 }

clang/test/Sema/__try.c

 
 void TEST(void) {
   __try {
 
   }
 }  // expected-error{{expected '__except' or '__finally' block}}
 
 void TEST(void) {
-  __except (FilterExpression()) { // expected-warning{{implicit declaration of function '__except' is invalid in C99}} \
+  __except (FilterExpression()) { // expected-error{{implicit declaration of function '__except' is invalid in C99}} \
     // expected-error{{too few arguments to function call, expected 1, have 0}} \
     // expected-error{{expected ';' after expression}}
   }
 }
 
 void TEST(void) {
   __finally { } // expected-error{{}}
 }

clang/test/Sema/aarch64-tme-errors.c

-// RUN: %clang_cc1 -triple aarch64-eabi -verify %s
+// RUN: %clang_cc1 -triple aarch64-eabi -verify -std=c99 %s
 
 #include "arm_acle.h"
 
 void test_no_tme_funcs(void) {
   __tstart();         // expected-warning{{implicit declaration of function '__tstart'}}
   __builtin_tstart(); // expected-error{{use of unknown builtin '__builtin_tstart'}}
 }

clang/test/Sema/arm-no-fp16.c

 // RUN: %clang_cc1 -triple thumbv7-none-eabi %s -target-feature +neon \
 // RUN:   -fallow-half-arguments-and-returns -target-feature -fp16 \
-// RUN:   -fsyntax-only -verify
+// RUN:   -fsyntax-only -verify -std=c99
 
 // REQUIRES: aarch64-registered-target || arm-registered-target
 
 #include <arm_neon.h>
 
 float16x4_t test_vcvt_f16_f32(float32x4_t a) {
   return vcvt_f16_f32(a); // expected-warning{{implicit declaration of function 'vcvt_f16_f32'}}  expected-error{{returning 'int' from a function with incompatible result type 'float16x4_t'}}
 }

clang/test/Sema/bitfield.c

 typedef __typeof__(+(t5.n++)) Unsigned; // Post-increment is underspecified, but seems to
 typedef __typeof__(+(t5.n--)) Unsigned; // also act like compound-assignment.
 
 struct Test6 {
   : 0.0; // expected-error{{type name requires a specifier or qualifier}}
 };
 
 struct PR36157 {
-  int n : 1 ? 1 : implicitly_declare_function(); // expected-warning {{invalid in C99}}
+  int n : 1 ? 1 : implicitly_declare_function(); // expected-error {{invalid in C99}}
 };

clang/test/Sema/builtin-setjmp.c

 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=c,expected -DNO_JMP_BUF %s -ast-dump | FileCheck %s --check-prefixes=CHECK1,CHECK2
 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=c,expected -DWRONG_JMP_BUF %s -ast-dump | FileCheck %s --check-prefixes=CHECK1,CHECK2
 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=c,expected -DRIGHT_JMP_BUF %s -ast-dump | FileCheck %s --check-prefixes=CHECK1,CHECK2
 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=c,expected -DONLY_JMP_BUF %s -ast-dump | FileCheck %s --check-prefixes=CHECK1,CHECK2
-// RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=c,expected -DNO_SETJMP %s -ast-dump 2>&1 | FileCheck %s --check-prefixes=CHECK1,CHECK2
+// RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=c,expected -std=c99 -DNO_SETJMP %s -ast-dump 2>&1 | FileCheck %s --check-prefixes=CHECK1,CHECK2
 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=cxx,expected -x c++ -DNO_JMP_BUF %s -ast-dump | FileCheck %s --check-prefixes=CHECK1,CHECK2
 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=cxx,expected -x c++ -DWRONG_JMP_BUF %s -ast-dump | FileCheck %s --check-prefixes=CHECK1,CHECK2
 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=cxx,expected -x c++ -DRIGHT_JMP_BUF %s -ast-dump | FileCheck %s --check-prefixes=CHECK1,CHECK2
 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=cxx,expected -x c++ -DONLY_JMP_BUF %s -ast-dump | FileCheck %s --check-prefixes=CHECK2
 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify=cxx,expected -x c++ -DNO_SETJMP %s -ast-dump | FileCheck %s --check-prefixes=CHECK2
 
 #ifdef __cplusplus
 extern "C" {

clang/test/Sema/builtins.c

void test18(void) {
   result = __builtin___strlcat_chk(dst, src, sizeof(dst), sizeof(dst));
 
   ptr = __builtin___memccpy_chk(dst, src, '\037', sizeof(src));      // expected-error {{too few arguments to function call}}
   ptr = __builtin___strlcpy_chk(dst, src, sizeof(dst), sizeof(dst)); // expected-warning {{incompatible integer to pointer conversion}}
   ptr = __builtin___strlcat_chk(dst, src, sizeof(dst), sizeof(dst)); // expected-warning {{incompatible integer to pointer conversion}}
 }
 
 void no_ms_builtins(void) {
-  __assume(1); // expected-warning {{implicit declaration}}
-  __noop(1); // expected-warning {{implicit declaration}}
-  __debugbreak(); // expected-warning {{implicit declaration}}
+  __assume(1); // expected-error {{implicit declaration}}
+  __noop(1); // expected-error {{implicit declaration}}
+  __debugbreak(); // expected-error {{implicit declaration}}
 }
 
 void unavailable(void) {
   __builtin_operator_new(0); // expected-error {{'__builtin_operator_new' is only available in C++}}
   __builtin_operator_delete(0); // expected-error {{'__builtin_operator_delete' is only available in C++}}
 }
 
 // rdar://18259539

clang/test/Sema/cxx-as-c.c

 // RUN: %clang_cc1 %s -verify
 
 // PR36157
 struct Foo {
-  Foo(int n) : n_(n) {} // expected-error 1+{{}} expected-warning 1+{{}}
+  Foo(int n) : n_(n) {} // expected-error 1+{{}}
 private:
   int n;
 };
 int main() { Foo f; } // expected-error 1+{{}}

clang/test/Sema/implicit-builtin-decl.c

 void h() {
   int malloc(int); // expected-warning{{incompatible redeclaration of library function 'malloc'}}
   int strcpy(int); // expected-warning{{incompatible redeclaration of library function 'strcpy'}} \
   // expected-note{{'strcpy' is a builtin with type 'char *(char *, const char *)'}}
 }
 
 void f2() {
   fprintf(0, "foo"); // expected-warning{{declaration of built-in function 'fprintf' requires inclusion of the header <stdio.h>}} \
-   expected-warning {{implicit declaration of function 'fprintf' is invalid in C99}}
+   expected-error {{implicit declaration of function 'fprintf' is invalid in C99}}
 }
 
 // PR2892
 void __builtin_object_size(); // expected-error{{conflicting types}} \
 // expected-note{{'__builtin_object_size' is a builtin with type}}
 
 int a[10];
 

clang/test/Sema/implicit-decl.c

-// RUN: %clang_cc1 %s -verify -fsyntax-only -Werror=implicit-function-declaration
+// RUN: %clang_cc1 %s -verify -fsyntax-only -Werror=implicit-function-declaration -std=c99
+// RUN: %clang_cc1 %s -verify -fsyntax-only -std=c11
+// RUN: %clang_cc1 %s -verify=c2x -fsyntax-only -std=c2x
 
 /// -Werror-implicit-function-declaration is a deprecated alias used by many projects.
 // RUN: %clang_cc1 %s -verify -fsyntax-only -Werror-implicit-function-declaration
 
+// c2x-note@*:* {{'__builtin_va_list' declared here}}
+
 typedef int int32_t;
 typedef unsigned char Boolean;
 
-extern int printf(__const char *__restrict __format, ...); // expected-note{{'printf' declared here}}
+extern int printf(__const char *__restrict __format, ...); // expected-note{{'printf' declared here}} \
+                                                              c2x-note {{'printf' declared here}}
 
 void func(void) {
    int32_t *vector[16];
    const char compDesc[16 + 1];
    int32_t compCount = 0;
-   if (_CFCalendarDecomposeAbsoluteTimeV(compDesc, vector, compCount)) { // expected-error {{implicit declaration of function '_CFCalendarDecomposeAbsoluteTimeV' is invalid in C99}} expected-note {{previous implicit declaration}}
+   if (_CFCalendarDecomposeAbsoluteTimeV(compDesc, vector, compCount)) { // expected-error {{implicit declaration of function '_CFCalendarDecomposeAbsoluteTimeV' is invalid in C99}} \
+                                                                            expected-note {{previous implicit declaration}} \
+                                                                            c2x-error {{use of undeclared identifier '_CFCalendarDecomposeAbsoluteTimeV'}}
    }
 
    printg("Hello, World!\n"); // expected-error{{implicit declaration of function 'printg' is invalid in C99}} \
-                              // expected-note{{did you mean 'printf'?}}
+                                 expected-note{{did you mean 'printf'?}} \
+                                 c2x-error {{use of undeclared identifier 'printg'; did you mean 'printf'?}}
 
-  __builtin_is_les(1, 3); // expected-error{{use of unknown builtin '__builtin_is_les'}}
+  __builtin_is_les(1, 3); // expected-error{{use of unknown builtin '__builtin_is_les'}} \
+                             c2x-error {{unknown type name '__builtin_is_les'; did you mean '__builtin_va_list'?}} \
+                             c2x-error {{expected identifier or '('}} \
+                             c2x-error {{expected ')'}} \
+                             c2x-note {{to match this '('}}
 }
 Boolean _CFCalendarDecomposeAbsoluteTimeV(const char *componentDesc, int32_t **vector, int32_t count) { // expected-error {{conflicting types}}
  return 0;
 }
 
 
 // Test the typo-correction callback in Sema::ImplicitlyDefineFunction
 extern int sformatf(char *str, __const char *__restrict __format, ...); // expected-note{{'sformatf' declared here}}
 void test_implicit(void) {
-  int formats = 0;
+  int formats = 0; // c2x-note {{'formats' declared here}}
   formatd("Hello, World!\n"); // expected-error{{implicit declaration of function 'formatd' is invalid in C99}} \
-                              // expected-note{{did you mean 'sformatf'?}}
+                                 expected-note{{did you mean 'sformatf'?}} \
+                                 c2x-error {{use of undeclared identifier 'formatd'; did you mean 'formats'?}} \
+                                 c2x-error {{called object type 'int' is not a function or function pointer}}
 }

clang/test/Sema/implicit-ms-builtin-decl.c

 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify %s -fms-extensions
-// RUN: %clang_cc1 -triple i386-unknown-unknown -fsyntax-only -verify %s -fms-extensions
+// RUN: %clang_cc1 -triple i386-unknown-unknown -fsyntax-only -verify %s -fms-extensions -std=c99
 
 void f(void) {
   (void)_byteswap_ushort(42); // expected-warning{{implicitly declaring library function '_byteswap_ushort'}} \
   // expected-note{{include the header <stdlib.h> or explicitly provide a declaration for '_byteswap_ushort'}}
   (void)_byteswap_uint64(42LL); // expected-warning{{implicitly declaring library function '_byteswap_uint64'}} \
   // expected-note{{include the header <stdlib.h> or explicitly provide a declaration for '_byteswap_uint64'}}
 }
 

 void i(void) {
   (void)__mulh(21, 2);
   (void)__umulh(21, 2);
 }
 #endif
 
 #if defined(i386)
 void h(void) {
   (void)__mulh(21LL, 2LL);  // expected-warning{{implicit declaration of function '__mulh' is invalid}}
   (void)__umulh(21ULL, 2ULL);  // expected-warning{{implicit declaration of function '__umulh' is invalid}}
 }
 #endif

clang/test/Sema/typo-correction.c

 void banana(void);  // expected-note {{'banana' declared here}}
 int c11Generic(int arg) {
   _Generic(hello, int : banana)();  // expected-error-re {{use of undeclared identifier 'hello'{{$}}}}
   _Generic(arg, int : bandana)();  // expected-error {{use of undeclared identifier 'bandana'; did you mean 'banana'?}}
 }
 
 typedef long long __m128i __attribute__((__vector_size__(16)));
 int PR23101(__m128i __x) {
-  return foo((__v2di)__x);  // expected-warning {{implicit declaration of function 'foo'}} \
+  return foo((__v2di)__x);  // expected-error {{implicit declaration of function 'foo'}} \
                             // expected-error {{use of undeclared identifier '__v2di'}}
 }
 
 void f(long *a, long b) {
       __atomic_or_fetch(a, b, c);  // expected-error {{use of undeclared identifier 'c'}}
 }
 
 extern double cabs(_Complex double z);

clang/test/Sema/vla.c

-// RUN: %clang_cc1 %s -verify -fsyntax-only -pedantic
+// RUN: %clang_cc1 %s -verify -fsyntax-only -pedantic -std=c99
 
 int test1(void) {
   typedef int x[test1()];  // vla
   static int y = sizeof(x);  // expected-error {{not a compile-time constant}}
 }
 
 // PR2347
 void f (unsigned int m)

clang/test/Sema/warn-strict-prototypes.c

 // RUN: %clang_cc1 -triple i386-pc-unknown -fsyntax-only -Wstrict-prototypes -Wno-implicit-function-declaration -verify %s
-// RUN: %clang_cc1 -triple i386-pc-unknown -fsyntax-only -Wstrict-prototypes -fdiagnostics-parseable-fixits %s 2>&1 | FileCheck %s
+// RUN: %clang_cc1 -triple i386-pc-unknown -fsyntax-only -Wstrict-prototypes -Wno-implicit-function-declaration -fdiagnostics-parseable-fixits %s 2>&1 | FileCheck %s
 
 // function definition with 0 params, no prototype, no preceding declaration.
 void foo0() {} // expected-warning {{this old-style function definition is not preceded by a prototype}}
 
 // function declaration with unspecified params
 void foo1(); // expected-warning {{this function declaration is not a prototype}}
              // CHECK: fix-it:"{{.*}}":{[[@LINE-1]]:11-[[@LINE-1]]:11}:"void"
 // function declaration with 0 params

clang/test/VFS/module_missing_vfs.m

 // RUN: rm -rf %t && mkdir -p %t
 // RUN: echo "void funcA(void);" >> %t/a.h
 
-// RUN: not %clang_cc1 -fmodules -fimplicit-module-maps -fmodules-cache-path=%t/mcp -I %S/Inputs/MissingVFS %s -fsyntax-only -ivfsoverlay %t/vfs.yaml 2>&1 | FileCheck %s -check-prefix=ERROR
+// RUN: not %clang_cc1 -std=c99 -fmodules -fimplicit-module-maps -fmodules-cache-path=%t/mcp -I %S/Inputs/MissingVFS %s -fsyntax-only -ivfsoverlay %t/vfs.yaml 2>&1 | FileCheck %s -check-prefix=ERROR
 // ERROR: virtual filesystem overlay file '{{.*}}' not found
 // RUN: find %t/mcp -name "A-*.pcm" | count 1
 
 // RUN: sed -e "s@INPUT_DIR@%{/S:regex_replacement}/Inputs@g" -e "s@OUT_DIR@%{/t:regex_replacement}@g" %S/Inputs/MissingVFS/vfsoverlay.yaml > %t/vfs.yaml
-// RUN: %clang_cc1 -fmodules -fimplicit-module-maps -fmodules-cache-path=%t/mcp -I %S/Inputs/MissingVFS %s -fsyntax-only -ivfsoverlay %t/vfs.yaml
+// RUN: %clang_cc1 -std=c99 -fmodules -fimplicit-module-maps -fmodules-cache-path=%t/mcp -I %S/Inputs/MissingVFS %s -fsyntax-only -ivfsoverlay %t/vfs.yaml
 // RUN: find %t/mcp -name "A-*.pcm" | count 1
 
 @import A;
 void test(void) {
   funcA();
 }