Page MenuHomePhabricator

rovka (Diana Picus)
User

Projects

User does not belong to any projects.

User Details

User Since
Apr 21 2016, 3:27 AM (322 w, 6 d)

Recent Activity

Today

rovka updated the diff for D128261: [Fortran] Relax tolerance for FCVS tests.

Rebase.

Wed, Jun 29, 1:06 AM · Restricted Project, Restricted Project
rovka added a comment to D128261: [Fortran] Relax tolerance for FCVS tests.

Thanks for updating this!

Thanks for the explanation! I have updated the patch description with all the information. fpcmp does treat everything as double precision.

This makes sense to me. And indeed, fpcmp uses double for all floating point values.

relaxes the tolerance for the FCVS tests to permit both values, since they are both close enough (the absolute difference is small)

Yeah, the absolute difference is negligible. Note that FP_TOLERANCE is used as the relative tolerance when passed to fpcmp (see fpcmp invocation here and the help text from fpcmp documents its arguments). I think that 1e-7 as relative tolerance gives enough precision here. 1e-11 was too aggressive IMO - perhaps that was meant as absolute tolerance (which, btw, I don't see being specified)?

Wed, Jun 29, 1:04 AM · Restricted Project, Restricted Project
rovka accepted D128763: [flang][test] Remove RUN COMMANDS/EXPECTED OUTPUT/INPUT markers from test/Driver.

Missed a spot ;) Otherwise LGTM

Wed, Jun 29, 12:56 AM · Restricted Project, Restricted Project
rovka added a comment to D128260: [Fortran] Ignore whitespace in FCVS test results.

No change, trying to convince the pre-merge to apply the patch...

How confident are you that it should work? I mean, perhaps the pre-merge CI has not been configured correctly for llvm-test-suite? I've just tried arc patch D128260 and it worked fine for me locally. This looks like the pre-merge CI issue. Perhaps create a bug?

Wed, Jun 29, 12:49 AM · Restricted Project, Restricted Project
rovka committed rT8e4703af93b0: [Fortran] Ignore whitespace in FCVS test results (authored by rovka).
[Fortran] Ignore whitespace in FCVS test results
Wed, Jun 29, 12:37 AM · Restricted Project
rovka closed D128260: [Fortran] Ignore whitespace in FCVS test results.
Wed, Jun 29, 12:37 AM · Restricted Project, Restricted Project
rovka added a comment to D128260: [Fortran] Ignore whitespace in FCVS test results.

No change, trying to convince the pre-merge to apply the patch...

How confident are you that it should work? I mean, perhaps the pre-merge CI has not been configured correctly for llvm-test-suite? I've just tried arc patch D128260 and it worked fine for me locally. This looks like the pre-merge CI issue. Perhaps create a bug?

To me all this makes sense and hence approving as is, thanks! AFAIK, the only buildbots running these tests are managed by Linaro, so you will know if something is broken :)

Wed, Jun 29, 12:35 AM · Restricted Project, Restricted Project

Yesterday

rovka updated the diff for D128260: [Fortran] Ignore whitespace in FCVS test results.

No change, trying to convince the pre-merge to apply the patch...

Tue, Jun 28, 12:49 AM · Restricted Project, Restricted Project
rovka added inline comments to D128260: [Fortran] Ignore whitespace in FCVS test results.
Tue, Jun 28, 12:22 AM · Restricted Project, Restricted Project
rovka updated the diff for D128260: [Fortran] Ignore whitespace in FCVS test results.

Added more comments.
Hopefully the pre-merge CI will actually run this time...

Tue, Jun 28, 12:21 AM · Restricted Project, Restricted Project

Mon, Jun 27

rovka accepted D128043: [flang][driver] Add support for `-O{0|1|2|3}`.

Cool, thanks!

Mon, Jun 27, 2:03 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D128260: [Fortran] Ignore whitespace in FCVS test results.

The patch application failed.

Right, not sure why it doesn't find those files, I'll try a few things...

Mon, Jun 27, 2:00 AM · Restricted Project, Restricted Project
rovka updated the diff for D128260: [Fortran] Ignore whitespace in FCVS test results.

Try to fix CI patch application - take 2.

Mon, Jun 27, 1:50 AM · Restricted Project, Restricted Project
rovka updated the diff for D128260: [Fortran] Ignore whitespace in FCVS test results.

Try to fix CI patch application.

Mon, Jun 27, 1:33 AM · Restricted Project, Restricted Project
rovka set the repository for D128260: [Fortran] Ignore whitespace in FCVS test results to rT test-suite.
Mon, Jun 27, 1:14 AM · Restricted Project, Restricted Project

Fri, Jun 24

rovka added a comment to D128262: [Fortran] Avoid digits in character constant.

You can get list-directed output records to wrap at any length. The default is 79 but it can be overridden via FORT_FMT_RECL=n in the environment.

Fri, Jun 24, 12:26 AM · Restricted Project, Restricted Project
rovka added a comment to D128262: [Fortran] Avoid digits in character constant.

From what I managed to google up, gfortran never introduces a newline, but ifort does unless told otherwise on the command line. As you can see, flang also introduces a newline by default.

So, we know that the format of the output in this case is compiler-specific and the reference output was generated using gfortran. Generalizing it so that it works for other compilers makes sense to me.

think we should try to accept both. WDYT?

How can we achieve this? Wouldn't that require a dedicated "reference" file for every compiler that we want to support here? Also, by replacing digits with letters you are basically making fpcmp ignore this particular bit of output, right? So:

  • why not delete this particular test line if it's to be ignored anyway?
  • why is the generated output only really verified with fpcmp?
Fri, Jun 24, 12:22 AM · Restricted Project, Restricted Project

Thu, Jun 23

rovka added a comment to D128262: [Fortran] Avoid digits in character constant.

Hey Diana, cheers for looking into this!

Please bear with me, as this is not my are of expertise :)

Not mine either :D

Thu, Jun 23, 6:06 AM · Restricted Project, Restricted Project
rovka updated the diff for D128261: [Fortran] Relax tolerance for FCVS tests.

Relax the tolerance only for FM905 and FM907 (which are already special-cased in D128260).

Thu, Jun 23, 5:35 AM · Restricted Project, Restricted Project
rovka updated the diff for D128260: [Fortran] Ignore whitespace in FCVS test results.

Right, disable :)
Updated the patch to special-case FM905 and FM907. I think it doesn't hurt to leave a bit of list-directed output in the test suite just to make sure it doesn't randomly explode.

Thu, Jun 23, 5:34 AM · Restricted Project, Restricted Project
rovka added a comment to D128260: [Fortran] Ignore whitespace in FCVS test results.

One thing that we could do is to only enable whitespace for the 2 tests that use list-directed output. That's going to require a bit more effort in the CMakeLists.txt, but it would let us be strict about whitespace where specific formatting is used, and more relaxed where it isn't.

Thu, Jun 23, 2:32 AM · Restricted Project, Restricted Project
rovka added a comment to D128260: [Fortran] Ignore whitespace in FCVS test results.

@rovka Would you know what is the scope of the llvm-testsuite? Is it limited to flang, or is it flang+gfortran, or is it any fortran compiler? And is it possible to have multiple outputs for comparison as @clementval mentioned?

Thu, Jun 23, 2:06 AM · Restricted Project, Restricted Project
rovka added a comment to D128043: [flang][driver] Add support for `-O{0|1|2|3}`.

I just realized I haven't pestered you enough about testing :) Can you add a test that -O4 indeed warns and uses -O3? Also, the summary says this should work in both the compilation and the frontend driver, but you're only testing with %flang_fc1.

Thu, Jun 23, 1:45 AM · Restricted Project, Restricted Project, Restricted Project
rovka updated the summary of D128260: [Fortran] Ignore whitespace in FCVS test results.
Thu, Jun 23, 1:20 AM · Restricted Project, Restricted Project

Wed, Jun 22

rovka added inline comments to D128043: [flang][driver] Add support for `-O{0|1|2|3}`.
Wed, Jun 22, 5:41 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D128260: [Fortran] Ignore whitespace in FCVS test results.

Looks OK. Assuming there are no tests that explicitly check for spacing. What is FPCMP usually set to?

Wed, Jun 22, 4:32 AM · Restricted Project, Restricted Project
rovka added a comment to D128262: [Fortran] Avoid digits in character constant.

How does fpcmp fit into this issue? Is that in the checker?

Wed, Jun 22, 4:28 AM · Restricted Project, Restricted Project
rovka added a comment to D128261: [Fortran] Relax tolerance for FCVS tests.

Thanks for the explanation! I have updated the patch description with all the information. fpcmp does treat everything as double precision.

Wed, Jun 22, 3:09 AM · Restricted Project, Restricted Project
rovka updated the summary of D128261: [Fortran] Relax tolerance for FCVS tests.
Wed, Jun 22, 3:01 AM · Restricted Project, Restricted Project

Tue, Jun 21

rovka accepted D126164: [flang][Driver] Refine _when_ driver diagnostics are formatted.

Yep, LGTM, thanks for the tests!
(Disclaimer: I don't consider myself a Clang developer, so maybe wait a day or 2 before committing in case someone else has comments)

Tue, Jun 21, 6:08 AM · Restricted Project, Restricted Project, Restricted Project
rovka requested review of D128269: Teach fpcmp about the x. FP format.
Tue, Jun 21, 5:59 AM · Restricted Project, Restricted Project
rovka accepted D128050: [flang][driver] Use `-O{0|1|2|3}` to define LLVM backend pass pipeline.

Seems fine, thanks.

Tue, Jun 21, 5:46 AM · Restricted Project, Restricted Project
rovka requested review of D128262: [Fortran] Avoid digits in character constant.
Tue, Jun 21, 5:20 AM · Restricted Project, Restricted Project
rovka requested review of D128261: [Fortran] Relax tolerance for FCVS tests.
Tue, Jun 21, 5:17 AM · Restricted Project, Restricted Project
rovka requested review of D128260: [Fortran] Ignore whitespace in FCVS test results.
Tue, Jun 21, 5:14 AM · Restricted Project, Restricted Project
rovka added a comment to D126164: [flang][Driver] Refine _when_ driver diagnostics are formatted.

Hi Peixin,

Tue, Jun 21, 12:20 AM · Restricted Project, Restricted Project, Restricted Project

Mon, Jun 20

rovka committed rG26041e17006c: Update link job for flang on windows (authored by rovka).
Update link job for flang on windows
Mon, Jun 20, 12:34 AM · Restricted Project, Restricted Project, Restricted Project
rovka closed D126291: [flang][Driver] Update link job on windows.
Mon, Jun 20, 12:33 AM · Restricted Project, Restricted Project, Restricted Project

Fri, Jun 17

rovka added a comment to D126291: [flang][Driver] Update link job on windows.

I'm going to go ahead and commit this on Monday if nobody raises any objections until then. Thanks again to everyone for all the help & feedback!

Fri, Jun 17, 1:05 AM · Restricted Project, Restricted Project, Restricted Project

Thu, Jun 16

rovka committed rG24b98520e2e8: Update FileCheck docs after D95849. NFCI (authored by rovka).
Update FileCheck docs after D95849. NFCI
Thu, Jun 16, 1:19 AM · Restricted Project, Restricted Project
rovka closed D127846: Update FileCheck docs after D95849. NFC.
Thu, Jun 16, 1:18 AM · Restricted Project, Restricted Project

Wed, Jun 15

rovka requested review of D127846: Update FileCheck docs after D95849. NFC.
Wed, Jun 15, 4:31 AM · Restricted Project, Restricted Project
rovka added inline comments to D126291: [flang][Driver] Update link job on windows.
Wed, Jun 15, 12:28 AM · Restricted Project, Restricted Project, Restricted Project

Tue, Jun 14

rovka updated the diff for D126291: [flang][Driver] Update link job on windows.
  • Use --implicit-not-check <3
Tue, Jun 14, 3:16 AM · Restricted Project, Restricted Project, Restricted Project
rovka added inline comments to D126291: [flang][Driver] Update link job on windows.
Tue, Jun 14, 3:05 AM · Restricted Project, Restricted Project, Restricted Project
rovka added inline comments to D126291: [flang][Driver] Update link job on windows.
Tue, Jun 14, 1:21 AM · Restricted Project, Restricted Project, Restricted Project
rovka updated the diff for D126291: [flang][Driver] Update link job on windows.

Clarified test checks.

Tue, Jun 14, 1:01 AM · Restricted Project, Restricted Project, Restricted Project

Mon, Jun 13

rovka updated the diff for D126291: [flang][Driver] Update link job on windows.
  • Stop using --ld-path, since it seems to be ignored on Windows. Instead, just match any path that ends in ld (should work with ld, lld, gold). This isn't very robust, but I don't have any better ideas (other than actually support --ld-path everywhere, not sure how much work that would be).
Mon, Jun 13, 4:26 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D126291: [flang][Driver] Update link job on windows.

I'm guessing the Windows precommit is failing because --ld-path is ignored on Windows, even if we use a Linux target? I have a fix that works on my Windows machine, coming right up.

Mon, Jun 13, 4:15 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D126291: [flang][Driver] Update link job on windows.

I had the same idea about switching the tests to using target triples instead of having separate files for it, but initially I had some issues getting that to work properly. When specifying a triple, we need to provide an architecture. Leaving the triple as unkown-linux-gnu or just linux-gnu gives us an error along the lines of flang-new: error: unknown target triple 'unknown-unknown-linux-gnu', please use -triple or -arch. OTOH, hardcoding an architecture like x86 or aarch64 fails if we're not building that specific backend.

If you actually execute code generation, then yes, it fails if that specific arch isn't enabled. But for general compiler driver level tests, which just print out the command the would have executed (when running with -###), it should work without the actual code generation target being available. This is at least how it's done for clang's corresponding tests, most files in clang/test/Driver have hardcoded arch triples, without any REQUIRES lines or other exclusions.

Mon, Jun 13, 3:08 AM · Restricted Project, Restricted Project, Restricted Project
rovka added reviewers for D127455: [MLIR] Add `comdat` to globals when translating to LLVM IR on Windows: mehdi_amini, rriddle.
Mon, Jun 13, 2:36 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D126291: [flang][Driver] Update link job on windows.

Moved the check for -flang-experimental-exec into addFortranRuntimeLibraryPath, so it affects all the toolchains. @awarzynski does this look like a good idea?

If moving that check to inside that function is ok, should the same check be added to addFortranRuntimeLibs, too?
Edit: And also retain that condition for any flags that are added in the respective part of the toolchain files that don't use any of these two functions?

Mon, Jun 13, 2:18 AM · Restricted Project, Restricted Project, Restricted Project
rovka updated the diff for D126291: [flang][Driver] Update link job on windows.

Missed a spot (removing the namespace in MinGW.cpp)

Mon, Jun 13, 2:12 AM · Restricted Project, Restricted Project, Restricted Project
rovka updated the diff for D126291: [flang][Driver] Update link job on windows.
  • Switch to the new style of testing (using triples rather than separate files)
  • Fix oversight in Darwin toolchain file
Mon, Jun 13, 2:09 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D126291: [flang][Driver] Update link job on windows.

I had the same idea about switching the tests to using target triples instead of having separate files for it, but initially I had some issues getting that to work properly. When specifying a triple, we need to provide an architecture. Leaving the triple as unkown-linux-gnu or just linux-gnu gives us an error along the lines of flang-new: error: unknown target triple 'unknown-unknown-linux-gnu', please use -triple or -arch. OTOH, hardcoding an architecture like x86 or aarch64 fails if we're not building that specific backend. We could do that and make the test REQUIRE the architecture that we're hardcoding, but this isn't really an architecture-specific test. So what I've finally done instead is to check for flang supported architectures and add a lit substitution for the first one that we find (be it aarch64, powerpc or x86) and use that in the test. We'll still get an error if someone tries to build the test without enabling any of these targets, but I think that's a good thing, since then people can decide either to add their architecture to lit or just, you know, not build flang on platforms where it isn't supported :) Patch incoming.

Mon, Jun 13, 2:07 AM · Restricted Project, Restricted Project, Restricted Project

Fri, Jun 10

rovka committed rG46f08a4ee0e9: [flang] Fix RUN line in test from 3e782ba (authored by rovka).
[flang] Fix RUN line in test from 3e782ba
Fri, Jun 10, 4:39 AM · Restricted Project, Restricted Project
rovka committed rG99fe38a13a2d: [flang][runtime] Remove dependency on C++ <mutex> on Windows (authored by mmuetzel).
[flang][runtime] Remove dependency on C++ <mutex> on Windows
Fri, Jun 10, 4:29 AM · Restricted Project, Restricted Project
rovka closed D127316: [flang][runtime] Remove dependency on C++ <mutex> on Windows.
Fri, Jun 10, 4:29 AM · Restricted Project, Restricted Project
rovka committed rT25c31a4e8b54: [Fortran] Don't add -Werror=date-time for flang (authored by rovka).
[Fortran] Don't add -Werror=date-time for flang
Fri, Jun 10, 2:25 AM · Restricted Project
rovka closed D126951: [Fortran] Don't add -Werror=date-time for flang.
Fri, Jun 10, 2:25 AM · Restricted Project, Restricted Project
rovka updated the diff for D126291: [flang][Driver] Update link job on windows.

Moved the check for -flang-experimental-exec into addFortranRuntimeLibraryPath, so it affects all the toolchains. @awarzynski does this look like a good idea?

Fri, Jun 10, 2:11 AM · Restricted Project, Restricted Project, Restricted Project
rovka added inline comments to D126291: [flang][Driver] Update link job on windows.
Fri, Jun 10, 12:55 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D127316: [flang][runtime] Remove dependency on C++ <mutex> on Windows.

Thanks for submitting this! I think the NOMINMAX trick should fix the issues I was seeing on Windows on Arm, but I still need to test to be sure (it's slow...). Do you have commit access or should I commit on your behalf?
One small Phabricator nitpick: it's nice to upload patches with context (you don't have to do that for this patch, I was just mentioning it for the future).

Fri, Jun 10, 12:49 AM · Restricted Project, Restricted Project

Wed, Jun 8

rovka updated the diff for D126291: [flang][Driver] Update link job on windows.
  • Fixed test
  • Unconditionally added the subsystem arg
  • Incorporated the MinGW toolchain changes (Thanks again @mmuetzel, I'm adding you as co-author)
Wed, Jun 8, 7:26 AM · Restricted Project, Restricted Project, Restricted Project
rovka accepted D126890: [flang][driver] Make `flang-new -fc1` accept MLIR files.

This LGTM, but please wait another day or two before committing in case anyone else still has comments.

Wed, Jun 8, 3:43 AM · Restricted Project, Restricted Project
rovka accepted D127207: [flang][driver] Fix support for `-x`.

LGTM, thanks.

Wed, Jun 8, 3:35 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D126955: [flang] Add RUN lines using `fir-opt`.

Thanks, this LGTM but I'll let @clementval have the final word :)

Wed, Jun 8, 3:11 AM · Restricted Project, Restricted Project

Fri, Jun 3

rovka added a comment to D126291: [flang][Driver] Update link job on windows.

@mmuetzel Thanks for looking into this! I think since you're passing -DCLANG_DEFAULT_RTLIB=compiler-rt, you might indeed need to build compiler-rt (or at least the builtins part of it).

Fri, Jun 3, 6:22 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D126890: [flang][driver] Make `flang-new -fc1` accept MLIR files.

This looks good in general, but it's missing a couple of tests:

  • a test with a file with a .mlir extension
  • tests for the -x option
Fri, Jun 3, 2:52 AM · Restricted Project, Restricted Project
rovka updated the diff for D126291: [flang][Driver] Update link job on windows.
  • Update for MinGW.
  • Add /subsystem:console to help link.exe understand what's going on.
Fri, Jun 3, 2:36 AM · Restricted Project, Restricted Project, Restricted Project
rovka closed D126690: [flang] Upstream lowering of real control loops.

Committed as https://github.com/llvm/llvm-project/commit/a159128811ae1d0a781dbc8590fa1e0f26642f82

Fri, Jun 3, 2:21 AM · Restricted Project, Restricted Project
rovka requested review of D126951: [Fortran] Don't add -Werror=date-time for flang.
Fri, Jun 3, 2:07 AM · Restricted Project, Restricted Project
rovka committed rTf425709b9cd0: [Fortran] Forward NO_STOP_MESSAGE to FCVS tests (authored by rovka).
[Fortran] Forward NO_STOP_MESSAGE to FCVS tests
Fri, Jun 3, 1:22 AM · Restricted Project
rovka closed D126360: [Fortran] Forward NO_STOP_MESSAGE to FCVS tests.
Fri, Jun 3, 1:22 AM · Restricted Project, Restricted Project
rovka committed rG7ce8c59b383a: [flang][test-suite] Document need for NO_STOP_MESSAGE environment variable. NFC (authored by rovka).
[flang][test-suite] Document need for NO_STOP_MESSAGE environment variable. NFC
Fri, Jun 3, 1:19 AM · Restricted Project, Restricted Project

Thu, Jun 2

rovka accepted D126712: Document CMake support.

Yay :)

Thu, Jun 2, 3:58 AM · Restricted Project, Restricted Project
rovka added inline comments to D126291: [flang][Driver] Update link job on windows.
Thu, Jun 2, 12:22 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D126291: [flang][Driver] Update link job on windows.

While the test passes now, I still get LINK : fatal error LNK1561: entry point must be defined when trying to actually link. Isn't it expected to not work yet?

Linking with lld-link.exe does work. objdump says there is a _QQmain symbol, why does lld consider this to be the entry point?

Thu, Jun 2, 12:17 AM · Restricted Project, Restricted Project, Restricted Project

Wed, Jun 1

rovka accepted D126697: [Flang] Lower the infinite do loop.

LGTM, thanks.

Wed, Jun 1, 3:56 AM · Restricted Project, Restricted Project
rovka updated the diff for D126291: [flang][Driver] Update link job on windows.

Updated comment in linker-flags.f90 test

Wed, Jun 1, 1:38 AM · Restricted Project, Restricted Project, Restricted Project
rovka added a comment to D126291: [flang][Driver] Update link job on windows.

With this patch in place, we still need to add `-Xlinker -subsystem:console' in order to compile a simple 'Hello World' to run from the console.

Is this behavior consistent with "clang.exe"? Also, any clue why is this needed? Some documentation here. I've scanned it and am no wiser :/

Wed, Jun 1, 1:37 AM · Restricted Project, Restricted Project, Restricted Project
rovka committed rGa159128811ae: [flang] Upstream lowering of real control loops (authored by rovka).
[flang] Upstream lowering of real control loops
Wed, Jun 1, 1:01 AM · Restricted Project, Restricted Project

Tue, May 31

rovka updated the summary of D126291: [flang][Driver] Update link job on windows.
Tue, May 31, 5:09 AM · Restricted Project, Restricted Project, Restricted Project
rovka added inline comments to D126291: [flang][Driver] Update link job on windows.
Tue, May 31, 5:08 AM · Restricted Project, Restricted Project, Restricted Project
rovka updated the diff for D126291: [flang][Driver] Update link job on windows.

Updated test to check for link.exe instead of lld-link.exe. This doesn't look great but it works for both lld-link.exe and the default link.exe. I tried to use --ld-path=various/paths instead, but it seems to be ignored on Windows (I even get a warning about "argument unused during compilation"). If anyone has any ideas about how to make the test more robust, please let me know.

Tue, May 31, 5:06 AM · Restricted Project, Restricted Project, Restricted Project
rovka requested review of D126690: [flang] Upstream lowering of real control loops.
Tue, May 31, 4:23 AM · Restricted Project, Restricted Project
rovka committed rG11fb1aa5a408: [flang] Upstream the lowering of the while loop (authored by rovka).
[flang] Upstream the lowering of the while loop
Tue, May 31, 12:28 AM · Restricted Project, Restricted Project
rovka closed D126636: [Flang] Upstream the lowering of the while loop.
Tue, May 31, 12:27 AM · Restricted Project, Restricted Project

May 30 2022

rovka requested review of D126636: [Flang] Upstream the lowering of the while loop.
May 30 2022, 1:29 AM · Restricted Project, Restricted Project

May 27 2022

rovka added a comment to D126279: [Fortran] Set QUIET = .TRUE. in STOP statements.

FCVS is very old code; it antedates the recent STOP QUIET feature. We run it because Fortran is short on publicly available test suites and because we want to be able to say that we pass this old famous suite. I agree that it would be best if FCVS's official sources could be updated, but it's basically a museum piece now.

May 27 2022, 2:42 AM · Restricted Project, Restricted Project
rovka added a comment to D126291: [flang][Driver] Update link job on windows.

@rovka in case you missed this, the test is failing in windows.

May 27 2022, 2:31 AM · Restricted Project, Restricted Project, Restricted Project

May 25 2022

rovka added a comment to D126360: [Fortran] Forward NO_STOP_MESSAGE to FCVS tests.

This is an alternative to https://reviews.llvm.org/D126279
I personally prefer the other patch, for reasons which I've explained there.
If we go with this approach, I'll also send a patch to the docs to document the need for setting this environment variable.

May 25 2022, 2:25 AM · Restricted Project, Restricted Project
rovka added a comment to D126279: [Fortran] Set QUIET = .TRUE. in STOP statements.

I can do that, but I don't like that solution. My understanding is that these warnings are platform-dependent, and also the 'Fortran STOP' seems to be printed by default by flang. This means pretty much anyone running the test-suite with flang will have to set this environment variable, or else face a ton of failures. Why make the test-suite harder to run? I think one of the goals of the test-suite is to be independent from both the hardware and the compiler, and all the better if it's possible to achieve that without compiler-specific hacks. In this case, setting the STOP statement to QUIET makes sense to me because its output is not strictly defined by the standard and will vary with both hardware and compiler. The test-suite is not a good place to test warnings, error messages or other fuzzy output - that should go in flang's own tests.

May 25 2022, 2:23 AM · Restricted Project, Restricted Project
rovka requested review of D126360: [Fortran] Forward NO_STOP_MESSAGE to FCVS tests.
May 25 2022, 2:06 AM · Restricted Project, Restricted Project

May 24 2022

rovka requested review of D126291: [flang][Driver] Update link job on windows.
May 24 2022, 4:38 AM · Restricted Project, Restricted Project, Restricted Project
rovka requested review of D126279: [Fortran] Set QUIET = .TRUE. in STOP statements.
May 24 2022, 1:15 AM · Restricted Project, Restricted Project

May 19 2022

rovka accepted D125957: [flang][driver] Make driver accept `-module-dir<value>`.

Seems legit, thank you!

May 19 2022, 2:53 AM · Restricted Project, Restricted Project, Restricted Project

May 18 2022

rovka committed rG1c0b03f6e706: [flang][driver] Support parsing response files (authored by rovka).
[flang][driver] Support parsing response files
May 18 2022, 1:26 AM · Restricted Project, Restricted Project
rovka closed D124846: [flang][driver] Support reading response files.
May 18 2022, 1:26 AM · Restricted Project, Restricted Project
rovka committed rG6bcafce103a4: [flang][Runtime] Use proper prototypes in Fortran_main. NFCI (authored by rovka).
[flang][Runtime] Use proper prototypes in Fortran_main. NFCI
May 18 2022, 1:04 AM · Restricted Project, Restricted Project