lld-link, clang: Treat non-existent input files as possible spellos for option flags

[thakis]

Summary

OptTable treats arguments starting with / that aren't a known option
as filenames. This means lld-link's and clang-cl's typo correction for
unknown flags didn't do spell checking for misspelled options that start
with /.

I first tried changing OptTable, but that got pretty messy, see PR41787
comments 2 and 3.

Instead, let lld-link's and clang's (including clang-cl's) "file not
found" diagnostic check if a non-existent file looks like it could be a
mis-spelled option, and if so add a "did you mean" suggestion to the
"file not found" diagnostic.

While here, make formatting of a few diagnostics a bit more
self-consistent.

Fixes PR41787.

[hans]

Beautiful!

Want to add a line to docs/ReleaseNotes.rst too? :-)

clang/include/clang/Basic/DiagnosticDriverKinds.td

177 ↗

(On Diff #200845)

(grammar nit: I think this is a comma splice (https://en.wikipedia.org/wiki/Comma_splice) so I believe a semicolon would be better. On the other hand, we seem to do this a lot already and I'm not a native speaker so maybe it's fine.)

clang/include/clang/Driver/Driver.h

398 ↗

(On Diff #200845)

Should the comment say what TypoCorrect does?

clang/lib/Driver/Driver.cpp

2024 ↗

(On Diff #200845)

indentation looks funny; clang-format?

lld/COFF/Driver.cpp

218 ↗

(On Diff #200845)

I like the semicolon better, but it seems we use comma in other places?

[thakis]

Thanks! Landing with comments addressed.

clang/include/clang/Basic/DiagnosticDriverKinds.td

177 ↗

(On Diff #200845)

Looking through rg 'did you mean' clang/include/clang/Basic/ we're actually pretty consistent about using a semicolon everywhere except for this file, so changed to ;. Thanks!

clang/include/clang/Driver/Driver.h

398 ↗

(On Diff #200845)

Sure, added.

lld/COFF/Driver.cpp

218 ↗

(On Diff #200845)

This is a semicolon because lld-link includes the strerror output (.second.message()), so this looks like could not open 'foo': No such file; did you mean '/foo'. Without the semicolon it looks like the "did you mean" is part of the error message.

[rnk]

Thanks, this has been a longstanding issue!

COFF/Driver.cpp

 }
 
 void LinkerDriver::enqueuePath(StringRef Path, bool WholeArchive) {
   auto Future =
       std::make_shared<std::future<MBErrPair>>(createFutureForFile(Path));
   std::string PathStr = Path;
   enqueueTask([=]() {
     auto MBOrErr = Future->get();
-    if (MBOrErr.second)
-      error("could not open " + PathStr + ": " + MBOrErr.second.message());
+    if (MBOrErr.second) {
+      std::string Error =
+          "could not open '" + PathStr + "': " + MBOrErr.second.message();
+      // Check if the filename is a typo for an option flag. OptTable thinks
+      // that all args that are not known options and that start with / are
+      // filenames, but e.g. `/nodefaultlibs` is more likely a typo for
+      // the option `/nodefaultlib` than a reference to a file in the root
+      // directory.
+      std::string Nearest;
+      if (COFFOptTable().findNearest(PathStr, Nearest) > 1)
+        error(Error);
       else
+        error(Error + "; did you mean '" + Nearest + "'");
+    } else
       Driver->addBuffer(std::move(MBOrErr.first), WholeArchive);
   });
 }
 
 void LinkerDriver::addArchiveBuffer(MemoryBufferRef MB, StringRef SymName,
                                     StringRef ParentName,
                                     uint64_t OffsetInArchive) {
   file_magic Magic = identify_magic(MB.getBuffer());

COFF/DriverUtils.cpp

opt::InputArgList ArgParser::parse(ArrayRef<const char *> Argv) {
 
   // We need to get the quoting style for response files before parsing all
   // options so we parse here before and ignore all the options but
   // --rsp-quoting.
   opt::InputArgList Args = Table.ParseArgs(Argv, MissingIndex, MissingCount);
 
   // Expand response files (arguments in the form of @<filename>)
   // and then parse the argument again.
-  SmallVector<const char *, 256> ExpandedArgv(Argv.data(), Argv.data() + Argv.size());
+  SmallVector<const char *, 256> ExpandedArgv(Argv.data(),
+                                              Argv.data() + Argv.size());
   cl::ExpandResponseFiles(Saver, getQuotingStyle(Args), ExpandedArgv);
   Args = Table.ParseArgs(makeArrayRef(ExpandedArgv).drop_front(), MissingIndex,
                          MissingCount);
 
   // Print the real command line if response files are expanded.
   if (Args.hasArg(OPT_verbose) && Argv.size() != ExpandedArgv.size()) {
     std::string Msg = "Command line:";
     for (const char *S : ExpandedArgv)

test/COFF/color-diagnostics.test

 # Windows command prompt doesn't support ANSI escape sequences.
 # REQUIRES: shell
 
 # RUN: not lld-link -xyz --color-diagnostics /nosuchfile 2>&1 \
 # RUN:   | FileCheck -check-prefix=COLOR %s
 # RUN: not lld-link -xyz --color-diagnostics=always /nosuchfile 2>&1 \
 # RUN:   | FileCheck -check-prefix=COLOR %s
 
 # COLOR: {{lld-link: .\[0;1;35mwarning: .\[0mignoring unknown argument '-xyz'}}
-# COLOR: {{lld-link: .\[0;1;31merror: .\[0mcould not open /nosuchfile}}
+# COLOR: {{lld-link: .\[0;1;31merror: .\[0mcould not open '/nosuchfile'}}
 
 # RUN: not lld-link /nosuchfile 2>&1 | FileCheck -check-prefix=NOCOLOR %s
 # RUN: not lld-link -color-diagnostics=never /nosuchfile 2>&1 \
 # RUN:   | FileCheck -check-prefix=NOCOLOR %s
 # RUN: not lld-link -color-diagnostics=always -no-color-diagnostics \
 # RUN:   /nosuchfile 2>&1 | FileCheck -check-prefix=NOCOLOR %s
 
-# NOCOLOR: lld-link: error: could not open /nosuchfile
+# NOCOLOR: lld-link: error: could not open '/nosuchfile'

test/COFF/could-not-open.test

 RUN: not lld-link 01 2>&1 | FileCheck %s
 
-CHECK:     could not open 01
+CHECK:     could not open '01'
 CHECK-NOT: /machine is not specified
 CHECK-NOT: subsystem must be defined

test/COFF/driver.test

 # RUN: not lld-link nosuchfile.obj >& %t.log
 # RUN: FileCheck -check-prefix=MISSING %s < %t.log
-MISSING: nosuchfile.obj: {{[Nn]}}o such file or directory
+MISSING: 'nosuchfile.obj': {{[Nn]}}o such file or directory
 
 # RUN: lld-link --version | FileCheck -check-prefix=VERSION %s
 VERSION: {{LLD [0-9]+\.[0-9]+}}
 
 # RUN: yaml2obj < %p/Inputs/export.yaml > %t.obj
 # RUN: lld-link /out:%t.dll /dll %t.obj
 # RUN: not lld-link /out:%t.exe %t.dll 2>&1 | FileCheck -check-prefix=BADFILE %s
 BADFILE: bad file type. Did you specify a DLL instead of an import library?

 
 # RUN: not lld-link -version 2>&1 | FileCheck -check-prefix=SPELLVERSION %s
 SPELLVERSION: ignoring unknown argument '-version', did you mean '--version'
 SPELLVERSION: no input files
 
 # RUN: not lld-link -nodefaultlibs 2>&1 | FileCheck -check-prefix=SPELLNODEFAULTLIB %s
 SPELLNODEFAULTLIB: ignoring unknown argument '-nodefaultlibs', did you mean '-nodefaultlib'
 SPELLNODEFAULTLIB: no input files
+
+# RUN: not lld-link /nodefaultlibs 2>&1 | FileCheck -check-prefix=SPELLNODEFAULTLIB_SLASH %s
+SPELLNODEFAULTLIB_SLASH: could not open '/nodefaultlibs': {{.*}}; did you mean '/nodefaultlib'
+SPELLNODEFAULTLIB_SLASH-NOT: no input files
+
+# Getting flags as typo corrections for normal input files is a side effect
+# of how spell checking for /-style flags is implemented.
+# RUN: not lld-link force 2>&1 | FileCheck -check-prefix=SPELLFORCE %s
+SPELLFORCE: could not open 'force': {{.*}}; did you mean '/force'
+SPELLFORCE-NOT: no input files

test/COFF/error-limit.test

 RUN: not lld-link 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 \
 RUN:   21 22 2>&1 | FileCheck -check-prefix=DEFAULT %s
 
-DEFAULT:      could not open 01
-DEFAULT:      could not open 20
+DEFAULT:      could not open '01'
+DEFAULT:      could not open '20'
 DEFAULT-NEXT: too many errors emitted, stopping now (use /errorlimit:0 to see all errors)
-DEFAULT-NOT:  could not open 21
+DEFAULT-NOT:  could not open '21'
 
 RUN: not lld-link /errorlimit:5 01 02 03 04 05 06 07 08 09 10 2>&1 \
 RUN:   | FileCheck -check-prefix=LIMIT5 %s
 
-LIMIT5:      could not open 01
-LIMIT5:      could not open 05
+LIMIT5:      could not open '01'
+LIMIT5:      could not open '05'
 LIMIT5-NEXT: too many errors emitted, stopping now (use /errorlimit:0 to see all errors)
-LIMIT5-NOT:  could not open 06
+LIMIT5-NOT:  could not open '06'
 
 RUN: not lld-link /errorlimit:0 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 \
 RUN:   16 17 18 19 20 21 22 2>&1 | FileCheck -check-prefix=UNLIMITED %s
 
-UNLIMITED:     could not open 01
-UNLIMITED:     could not open 20
-UNLIMITED:     could not open 21
-UNLIMITED:     could not open 22
+UNLIMITED:     could not open '01'
+UNLIMITED:     could not open '20'
+UNLIMITED:     could not open '21'
+UNLIMITED:     could not open '22'
 UNLIMITED-NOT: too many errors emitted, stopping now (use /errorlimit:0 to see all errors)
 
 RUN: not lld-link /errorlimit:XYZ 01 02 03 04 05 06 07 08 09 10 11 12 13 14 \
 RUN:   15 16 17 18 19 20 21 22 2>&1 | FileCheck -check-prefix=WRONG %s
 
 WRONG:      /errorlimit: number expected, but got XYZ

test/COFF/nodefaultlib.test

 # RUN:   /nodefaultlib:std64.lib >& %t.log || true
 # RUN: FileCheck -check-prefix=CHECK3 %s < %t.log
 
 # RUN: lld-link /libpath:%T /out:%t.exe /entry:main \
 # RUN:   /subsystem:console hello64.obj /defaultlib:std64 \
 # RUN:   /nodefaultlib:std64.lib >& %t.log || true
 # RUN: FileCheck -check-prefix=CHECK3 %s < %t.log
 
-CHECK1: error: could not open hello64.obj: {{[Nn]}}o such file or directory
-CHECK2: error: could not open hello64: {{[Nn]}}o such file or directory
+CHECK1: error: could not open 'hello64.obj': {{[Nn]}}o such file or directory
+CHECK2: error: could not open 'hello64': {{[Nn]}}o such file or directory
 CHECK3: error: undefined symbol: MessageBoxA
 CHECK3-NEXT: >>> referenced by {{.*}}hello64.obj:(main)
 
 # RUN: lld-link /libpath:%T /out:%t.exe /entry:main \
 # RUN:   /subsystem:console hello64.obj /defaultlib:std64.lib
 
 # RUN: env LIB=%T lld-link /out:%t.exe /entry:main \
 # RUN:   /subsystem:console hello64.obj /defaultlib:std64.lib

test/COFF/responsefile.test

 # RUN: yaml2obj < %p/Inputs/ret42.yaml > %t.obj
 
 # RUN: echo /out:%t.exe /entry:main %t.obj > %t.rsp
 # RUN: lld-link @%t.rsp /heap:0x3000
 # RUN: llvm-readobj --file-headers %t.exe | FileCheck %s
 CHECK: SizeOfHeapReserve: 12288
 
 # RUN: not lld-link --rsp-quoting=foobar @%t.rsp 2>&1 | \
 # RUN:     FileCheck --check-prefix=INVRSP %s
 INVRSP: invalid response file quoting: foobar
 
 # RUN: echo "blah\foo" > %t.rsp
 # RUN: not lld-link @%t.rsp 2>&1 | \
 # RUN:     FileCheck --check-prefix=DEFRSP %s
-DEFRSP: error: could not open blah\foo
+DEFRSP: error: could not open 'blah\foo'
 
 # RUN: echo "blah\foo" > %t.rsp
 # RUN: not lld-link --rsp-quoting=windows @%t.rsp 2>&1 | \
 # RUN:     FileCheck --check-prefix=WINRSP %s
-WINRSP: error: could not open blah\foo
+WINRSP: error: could not open 'blah\foo'
 
 # RUN: echo "blah\foo" > %t.rsp
 # RUN: not lld-link --rsp-quoting=posix @%t.rsp 2>&1 | \
 # RUN:     FileCheck --check-prefix=POSRSP %s
-POSRSP: error: could not open blahfoo
+POSRSP: error: could not open 'blahfoo'