tra added inline comments.
================
Comment at: clang/test/Driver/cuda-unsupported-debug-options.cu:18
+// RUN: %clang -### -target x86_64-linux-gnu -c %s -gdwarf-5 -gembed-source
2>&1 | FileCheck %s --check-prefix=DWARF-CLAMP
+// CHECK: debug information option
'{{-gz|-fdebug-info-for-profiling|-gsplit-dwarf|-glldb|-gcodeview|-gmodules|-gembed-source|-fdebug-macro|-ggnu-pubnames|-gdwarf-aranges|-fdebug-types-section}}'
is not supported{{.*}} target 'nvptx64-nvidia-cuda'
// CHECK-NOT: debug information option '{{.*}}' is not supported for target
'x86
----------------
dblaikie wrote:
> tra wrote:
> > dblaikie wrote:
> > > What's the purpose of the {{.*}} in this line?
> > The wildcard eats the differences between two warnings -- a generic one:
> > - `warning: debug information option '-glldb' is not supported for target
> > 'nvptx64-nvidia-cuda'`
> > and the more specific one for `-gembed-source` :
> > - `warning: debug information option '-gembed-source' is not supported. It
> > needs DWARF-5 but target 'nvptx64-nvidia-cuda' only provides DWARF-2.`
> >
> >
> Is that necessary? Or is the -gembed-source+CHECK test redundant with/less
> precise than the -gembed-source+DWARF-CLAMP test? (ie: would it be effective
> (not a loss in test coverage) to remove the gembed-source+CHECK test and only
> have the gembed-source+DWARF-CLAMP test, and keep the CHECK line precise as
> it was before for all the CHECK tests)
>
> Effectively: Rather than adding a new RUN line, and generalizing the CHECK
> over the new/old message and checking the new message separately - instead,
> change the existing gembed-source test to use DWARF-CLAMP instead of CHECK,
> and leave everything the same?
`CHECK` lines test that an option was ignored for the right target (nvptx, not
x86). So, both warning variants should participate.
`DWARF-CLAMP` verify that the correct DWARF version was used.
`-gembed-source` shoud participate in both tests. The options here is to either
include `-gembed-source` in the `CHECK` tests and make it accept both pattern
variants, or replicate the `target of the warning` check within DWARF-CLAMP.
I would prefer to keep these two kinds of tests separate.
I've just updated the patch and moved DWARF clamping tests into a separate file
and consolidated nearly-identical cuda/omp unsupported option warnings tests
into the same file.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D92617/new/
https://reviews.llvm.org/D92617
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
