[clang] Add -Wimplicit-fallthrough to -Wextra

[vegerot]

This patch adds -Wimplicit-fallthrough to -Wextra. GCC already includes it in -Wextra.

This patch also adds a test to check that -Wimplicit-fallthrough is included in -Wextra.

Note: This patch may regress performance when building with -Wextra.
This is because -Wextra requires forming a CFG for every function.

[github-actions]

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be
notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write
permissions for the repository. In which case you can instead tag reviewers by
name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review
by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate
is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

[llvmbot]

@llvm/pr-subscribers-libcxx

@llvm/pr-subscribers-clang

Author: Max Coplan (vegerot)

Changes

This patch adds -Wimplicit-fallthrough to -Wextra. GCC already includes it in -Wextra.

This patch also adds a test to check that -Wimplicit-fallthrough is included in -Wextra.

Note: This patch may regress performance when building with -Wextra.
This is because -Wextra requires forming a CFG for every function.

---

Full diff: https://github.com/llvm/llvm-project/pull/97926.diff

3 Files Affected:

• (modified) clang/include/clang/Basic/DiagnosticGroups.td (+1)
• (modified) clang/test/Sema/fallthrough-attr.c (+1)
• (modified) clang/test/SemaCXX/switch-implicit-fallthrough-macro.cpp (+1)

diff --git a/clang/include/clang/Basic/DiagnosticGroups.td b/clang/include/clang/Basic/DiagnosticGroups.td
index 2241f8481484e..a875f665b10f9 100644
--- a/clang/include/clang/Basic/DiagnosticGroups.td
+++ b/clang/include/clang/Basic/DiagnosticGroups.td
@@ -1063,6 +1063,7 @@ def Extra : DiagGroup<"extra", [
     StringConcatation,
     FUseLdPath,
     CastFunctionTypeMismatch,
+    ImplicitFallthrough,
   ]>;
 
 def Most : DiagGroup<"most", [
diff --git a/clang/test/Sema/fallthrough-attr.c b/clang/test/Sema/fallthrough-attr.c
index de50ebf39d42f..6cc19136f30a7 100644
--- a/clang/test/Sema/fallthrough-attr.c
+++ b/clang/test/Sema/fallthrough-attr.c
@@ -2,6 +2,7 @@
 // RUN: %clang_cc1 -fsyntax-only -std=gnu99 -verify -Wimplicit-fallthrough %s
 // RUN: %clang_cc1 -fsyntax-only -std=c99 -verify -Wimplicit-fallthrough %s
 // RUN: %clang_cc1 -fsyntax-only -std=c11 -verify -Wimplicit-fallthrough %s
+// RUN: %clang_cc1 -fsyntax-only -std=c11 -verify -Wextra %s
 // RUN: %clang_cc1 -fsyntax-only -std=c2x -DC2X -verify -Wimplicit-fallthrough %s
 
 int fallthrough_attribute_spelling(int n) {
diff --git a/clang/test/SemaCXX/switch-implicit-fallthrough-macro.cpp b/clang/test/SemaCXX/switch-implicit-fallthrough-macro.cpp
index 11df2cbfb53f0..cbbff1f1793b8 100644
--- a/clang/test/SemaCXX/switch-implicit-fallthrough-macro.cpp
+++ b/clang/test/SemaCXX/switch-implicit-fallthrough-macro.cpp
@@ -3,6 +3,7 @@
 // RUN: %clang_cc1 -fsyntax-only -verify -std=c++1z -Wimplicit-fallthrough -DCLANG_PREFIX -DCOMMAND_LINE_FALLTHROUGH=[[clang::fallthrough]] %s
 // RUN: %clang_cc1 -fsyntax-only -verify -std=c++1z -Wimplicit-fallthrough -DCOMMAND_LINE_FALLTHROUGH=[[clang::fallthrough]] %s
 // RUN: %clang_cc1 -fsyntax-only -verify -std=c++1z -Wimplicit-fallthrough -DCOMMAND_LINE_FALLTHROUGH=[[fallthrough]] -DUNCHOSEN=[[clang::fallthrough]] %s
+// RUN: %clang_cc1 -fsyntax-only -verify -std=c++1z -Wextra -DCOMMAND_LINE_FALLTHROUGH=[[fallthrough]] -DUNCHOSEN=[[clang::fallthrough]] %s
 
 int fallthrough_compatibility_macro_from_command_line(int n) {
   switch (n) {

[vegerot]

@AaronBallman please review

I don't know how to go about measuring if there even is a performance regression. How would I do that?

[AaronBallman]

@AaronBallman please review

I don't know how to go about measuring if there even is a performance regression. How would I do that?

I would put up a branch on http://llvm-compile-time-tracker.com (instructions in the About page) and see how this impacts compile times. We've always left diagnostics out of -Wextra that require a CFG, so getting performance measurements is important.

[vegerot]

I would put up a branch on http://llvm-compile-time-tracker.com (instructions in the About page) and see how this impacts compile times. We've always left diagnostics out of -Wextra that require a CFG, so getting performance measurements is important.

Thanks! The About page says

If you are an LLVM contributor who regularly does compile-time sensitive work

Do you think I should still ask for permission or should I leave it to someone else?

Update: I reached out to @nikic and CCd you

[nikic]

CTMark is not compiled with -Wextra, so we'd not get any useful data out of this patch. What changes does one have to do to enable -Wimplicit-fallthrough by default (not just for -Wextra)?

[vegerot]

CTMark is not compiled with -Wextra, so we'd not get any useful data out of this patch. What changes does one have to do to enable -Wimplicit-fallthrough by default (not just for -Wextra)?

@nikic done! At vegerot/llvm-project/perf/add-implicit-fallthrough-to-wextra

[nikic]

I'm not seeing any compile time impact from dropping the DefaultIgnores on warn_unannotated_fallthrough/warn_unannotated_fallthrough_per_function. So either it's free or I'm still doing something wrong. (http://llvm-compile-time-tracker.com/compare.php?from=6d12b3f67df429bffff6e1953d9f55867d7e2469&to=4117b087cc607686c472f54f82f1222688b5f2bf&stat=instructions:u)

[vegerot]

I'm pretty sure I enabled the warning correctly. @AaronBallman are we good to land?

[AaronBallman]

I'm not seeing any compile time impact from dropping the DefaultIgnores on warn_unannotated_fallthrough/warn_unannotated_fallthrough_per_function. So either it's free or I'm still doing something wrong. (http://llvm-compile-time-tracker.com/compare.php?from=6d12b3f67df429bffff6e1953d9f55867d7e2469&to=4117b087cc607686c472f54f82f1222688b5f2bf&stat=instructions:u)

Yeah, that logic should suffice:

  bool FallThroughDiagFull =
      !Diags.isIgnored(diag::warn_unannotated_fallthrough, D->getBeginLoc());
  bool FallThroughDiagPerFunction = !Diags.isIgnored(
      diag::warn_unannotated_fallthrough_per_function, D->getBeginLoc());
  if (FallThroughDiagFull || FallThroughDiagPerFunction ||
      fscope->HasFallthroughStmt) {
    DiagnoseSwitchLabelsFallthrough(S, AC, !FallThroughDiagFull);
  }

so unless something explicitly ignores the fallthrough warning in the setup, it seems like a valid test.

[AaronBallman]

I'm pretty sure I enabled the warning correctly. @AaronBallman are we good to land?

Not quite -- it looks like this change is causing libc++ test failures that were caught by precommit CI:

# | /var/lib/buildkite-agent/builds/linux-56-59b8f5d88-jltnz-1/llvm-project/github-pull-requests/build-runtimes/include/c++/v1/regex:3925:5: note: insert '[[clang::fallthrough]];' to silence this warning
# |  3925 |     case 'x':
# |       |     ^
# |       |     [[clang::fallthrough]];
# | /var/lib/buildkite-agent/builds/linux-56-59b8f5d88-jltnz-1/llvm-project/github-pull-requests/build-runtimes/include/c++/v1/regex:3925:5: note: insert 'break;' to avoid fall-through
# |  3925 |     case 'x':
# |       |     ^
# |       |     break;
# | 1 error generated.

CC @ldionne

[vegerot]

I'm pretty sure I enabled the warning correctly. @AaronBallman are we good to land?

Not quite -- it looks like this change is causing libc++ test failures that were caught by precommit CI:

Gotcha. Should I work on fixing the code?

[vegerot]

@AaronBallman I stacked this diff on top of #100821 to remove that build error.

question for reviewer: what is the proper etiquette for stacked diffs? Should I open two PRs or have one PR with two commits?

[vegerot]

@mordante @AaronBallman This diff is ready for review 🙂

[AaronBallman]

Sorry for the long wait on this review! A related issue cropped up regarding compile time performance of -Wall (#105959) and it seems to be caused by CFG creation.

I think we should run the test case from that issue against the changes in this patch (-Wall includes -Wextra) to see if it makes an even more noticeable change in compile times, just to be sure before we go ahead and land this. I was hopeful that CFG building got faster, but now I worry there may be edge cases that make CFG building particularly slow that we may need to investigate.

[vegerot]

@AaronBallman sure. Should I still use http://llvm-compile-time-tracker.com/ , or do I run the benchmark another way?

[AaronBallman]

@AaronBallman sure. Should I still use http://llvm-compile-time-tracker.com/ , or do I run the benchmark another way?

Sorry, your question completely fell off my radar! llvm-compile-time-tracker would be fine, but even just a local test would suffice -- basically, I expect the test case from #105959 to be slow before you apply your patch and am just wondering whether that gets even slower with your patch.