[cfe-commits] r154776 - in /cfe/trunk/test: Analysis/objc-bool.m Headers/typedef_guards.c Lexer/utf-16.c Preprocessor/pragma_sysheader.c Sema/surpress-deprecated.c SemaCXX/warn-everthing.cpp

[David Blaikie]

On Mon, Apr 16, 2012 at 11:03 AM, Chandler Carruth <chandlerc at google.com> wrote:
> On Mon, Apr 16, 2012 at 7:53 PM, David Blaikie <dblaikie at gmail.com> wrote:
>>
>> On Mon, Apr 16, 2012 at 10:43 AM, Chandler Carruth <chandlerc at google.com>
>> wrote:
>> > On Mon, Apr 16, 2012 at 7:38 PM, David Blaikie <dblaikie at gmail.com>
>> > wrote:
>> >>
>> >> On Mon, Apr 16, 2012 at 10:29 AM, Matthieu Monrocq
>> >> <matthieu.monrocq at gmail.com> wrote:
>> >> >
>> >> >
>> >> > Le 16 avril 2012 00:09, David Blaikie <dblaikie at gmail.com> a écrit :
>> >> >
>> >> >> Author: dblaikie
>> >> >> Date: Sun Apr 15 17:09:44 2012
>> >> >> New Revision: 154776
>> >> >>
>> >> >> URL: http://llvm.org/viewvc/llvm-project?rev=154776&view=rev
>> >> >> Log:
>> >> >> Fix tests that weren't actually verifying anything.
>> >> >>
>> >> >> Passing -verify to clang without -cc1 or -Xclang silently passes
>> >> >> (with
>> >> >> a
>> >> >> printed warning, but lit doesn't care about that). This change adds
>> >> >> -cc1
>> >> >> or,
>> >> >> as is necessary in one case, -Xclang to fix this so that these tests
>> >> >> are
>> >> >> actually verifying as intended.
>> >> >>
>> >> >> I'd like to change the driver so this kind of mistake could not be
>> >> >> made,
>> >> >> but
>> >> >> I'm not entirely sure how. Further, since the driver only warns
>> >> >> about
>> >> >> unknown
>> >> >> flags in general, we could have similar bugs with a misspellings of
>> >> >> arguments
>> >> >> that would be nice to find.
>> >> >>
>> >> >> Modified:
>> >> >>    cfe/trunk/test/Analysis/objc-bool.m
>> >> >>    cfe/trunk/test/Headers/typedef_guards.c
>> >> >>    cfe/trunk/test/Lexer/utf-16.c
>> >> >>    cfe/trunk/test/Preprocessor/pragma_sysheader.c
>> >> >>    cfe/trunk/test/Sema/surpress-deprecated.c
>> >> >>    cfe/trunk/test/SemaCXX/warn-everthing.cpp
>> >> >>
>> >> >> Modified: cfe/trunk/test/Analysis/objc-bool.m
>> >> >> URL:
>> >> >>
>> >> >>
>> >> >> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Analysis/objc-bool.m?rev=154776&r1=154775&r2=154776&view=diff
>> >> >>
>> >> >>
>> >> >>
>> >> >> ==============================================================================
>> >> >> --- cfe/trunk/test/Analysis/objc-bool.m (original)
>> >> >> +++ cfe/trunk/test/Analysis/objc-bool.m Sun Apr 15 17:09:44 2012
>> >> >> @@ -1,4 +1,4 @@
>> >> >> -// RUN: %clang --analyze %s -o %t -verify
>> >> >> +// RUN: %clang --analyze %s -o %t -Xclang -verify
>> >> >>
>> >> >>  // Test handling of ObjC bool literals.
>> >> >>
>> >> >>
>> >> >> Modified: cfe/trunk/test/Headers/typedef_guards.c
>> >> >> URL:
>> >> >>
>> >> >>
>> >> >> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Headers/typedef_guards.c?rev=154776&r1=154775&r2=154776&view=diff
>> >> >>
>> >> >>
>> >> >>
>> >> >> ==============================================================================
>> >> >> --- cfe/trunk/test/Headers/typedef_guards.c (original)
>> >> >> +++ cfe/trunk/test/Headers/typedef_guards.c Sun Apr 15 17:09:44 2012
>> >> >> @@ -1,4 +1,4 @@
>> >> >> -// RUN: %clang -fsyntax-only -verify %s
>> >> >> +// RUN: %clang_cc1 -fsyntax-only -verify %s
>> >> >>
>> >> >>  // NULL is rdefined in stddef.h
>> >> >>  #define NULL ((void*) 0)
>> >> >>
>> >> >> Modified: cfe/trunk/test/Lexer/utf-16.c
>> >> >> URL:
>> >> >>
>> >> >>
>> >> >> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Lexer/utf-16.c?rev=154776&r1=154775&r2=154776&view=diff
>> >> >>
>> >> >>
>> >> >>
>> >> >> ==============================================================================
>> >> >> --- cfe/trunk/test/Lexer/utf-16.c (original)
>> >> >> +++ cfe/trunk/test/Lexer/utf-16.c Sun Apr 15 17:09:44 2012
>> >> >> @@ -1,4 +1,4 @@
>> >> >> -// RUN: not %clang %s -fsyntax-only -verify
>> >> >> +// RUN: %clang_cc1 %s -fsyntax-only -verify
>> >> >>  // rdar://7876588
>> >> >>
>> >> >>  // This test verifies that clang gives a decent error for UTF-16
>> >> >> source
>> >> >> files.
>> >> >>
>> >> >> Modified: cfe/trunk/test/Preprocessor/pragma_sysheader.c
>> >> >> URL:
>> >> >>
>> >> >>
>> >> >> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Preprocessor/pragma_sysheader.c?rev=154776&r1=154775&r2=154776&view=diff
>> >> >>
>> >> >>
>> >> >>
>> >> >> ==============================================================================
>> >> >> --- cfe/trunk/test/Preprocessor/pragma_sysheader.c (original)
>> >> >> +++ cfe/trunk/test/Preprocessor/pragma_sysheader.c Sun Apr 15
>> >> >> 17:09:44
>> >> >> 2012
>> >> >> @@ -1,4 +1,4 @@
>> >> >> -// RUN: %clang -verify -pedantic %s -fsyntax-only
>> >> >> +// RUN: %clang_cc1 -verify -pedantic %s -fsyntax-only
>> >> >>  // RUN: %clang_cc1 -E %s | FileCheck %s
>> >> >>  // rdar://6899937
>> >> >>  #include "pragma_sysheader.h"
>> >> >>
>> >> >> Modified: cfe/trunk/test/Sema/surpress-deprecated.c
>> >> >> URL:
>> >> >>
>> >> >>
>> >> >> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Sema/surpress-deprecated.c?rev=154776&r1=154775&r2=154776&view=diff
>> >> >>
>> >> >>
>> >> >>
>> >> >> ==============================================================================
>> >> >> --- cfe/trunk/test/Sema/surpress-deprecated.c (original)
>> >> >> +++ cfe/trunk/test/Sema/surpress-deprecated.c Sun Apr 15 17:09:44
>> >> >> 2012
>> >> >> @@ -1,4 +1,4 @@
>> >> >> -// RUN: %clang -fsyntax-only -Wno-deprecated-declarations -verify
>> >> >> %s
>> >> >> +// RUN: %clang_cc1 -fsyntax-only -Wno-deprecated-declarations
>> >> >> -verify
>> >> >> %s
>> >> >>  extern void OldFunction() __attribute__((deprecated));
>> >> >>
>> >> >>  int main (int argc, const char * argv[]) {
>> >> >>
>> >> >> Modified: cfe/trunk/test/SemaCXX/warn-everthing.cpp
>> >> >> URL:
>> >> >>
>> >> >>
>> >> >> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/SemaCXX/warn-everthing.cpp?rev=154776&r1=154775&r2=154776&view=diff
>> >> >>
>> >> >>
>> >> >>
>> >> >> ==============================================================================
>> >> >> --- cfe/trunk/test/SemaCXX/warn-everthing.cpp (original)
>> >> >> +++ cfe/trunk/test/SemaCXX/warn-everthing.cpp Sun Apr 15 17:09:44
>> >> >> 2012
>> >> >> @@ -1,4 +1,4 @@
>> >> >> -// RUN: %clang -Weverything -fsyntax-only %s -verify
>> >> >> +// RUN: %clang_cc1 -Weverything -fsyntax-only %s -verify
>> >> >>
>> >> >>  // This previously crashed due to a bug in the CFG.  Exercising all
>> >> >>  // warnings helps check CFG construction.
>> >> >> @@ -8,6 +8,6 @@
>> >> >>   ~PR12271();
>> >> >>  };
>> >> >>
>> >> >> -void testPR12271() {
>> >> >> -  PR12271 a[1][1];
>> >> >> -}
>> >> >> \ No newline at end of file
>> >> >> +void testPR12271() { // expected-warning {{no previous prototype
>> >> >> for
>> >> >> function 'testPR12271'}}
>> >> >> +  PR12271 a[1][1]; // expected-warning {{unused variable 'a'}}
>> >> >> +}
>> >> >>
>> >> >>
>> >> >> _______________________________________________
>> >> >> cfe-commits mailing list
>> >> >> cfe-commits at cs.uiuc.edu
>> >> >> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>> >> >
>> >> >
>> >> > The only way I would see lit actually reacting would be to terminate
>> >> > the
>> >> > processus with a non-0 result, which should be reported.
>> >>
>> >> Right
>> >>
>> >> > Perhaps that we could add a "developer" option provoking this kind of
>> >> > behavior for unknown flags and have lit expand %clang and %clang_cc1
>> >> > automatically with this flag. This flag should then be negated in the
>> >> > few
>> >> > tests that actually test the driver diagnostic.
>> >>
>> >> Ideally, I'm wondering whether it makes sense to have even release
>> >> versions of clang produce a non-zero exit code for completely unknown
>> >> flags. If someone passes -foobar to clang, should we really silently
>> >> (wrt exit code, at least) succeed?
>> >
>> >
>> > What if the flag is a GCC flag that "might" work depending on what CC is
>> > set
>> > to?
>>
>> Are there particular cases of this that you're aware? For what it's
>> worth, GCC errors on unknown flags, which seems helpful. I could
>> imagine there being certain subcategories (-f*, -g*, etc) that we
>> could handle differently (either explicitly downgrading to a warning
>> for them, or whatever else seemed appropriate)
>
>
> Yes, there are loads of GCC flags that Clang doesn't support yet, even just
> to drop them on the floor.
>
> People expect Clang to be (essentially) a drop-in replacement for GCC. They
> don't expect the reverse (generally), although the fact that the reverse is
> not true has bitten us repeatedly already. I don't think we can justify
> Clang's behavior on GCC's here.
>
>>
>>
>> Actually, on further experimentation I don't know what GCC is doing.
>> -verify, -bar, -baz produce "unrecognized option" but don't fail.
>> "-foo" produces "cc1plus: error: unrecognized command line option
>> "-foo"" and fails.
>
>
> "-foo" is special -- it thinks it's an argument to cc1plus (-f... is
> generally an argument to cc1plus) which has a stricter behavior.
>
>>
>>
>> > I think we need to let users ask for either warnings or errors about
>> > unknown flags, just as we do today.
>>
>> So I understand you clearly - by this are you suggesting "no change",
>> or some kind of change that would still preserve that behavior?
>
>
> No change. Maybe I'm just not sufficiently paranoid, but this just doesn't
> seem like a big enough deal to expend this much effort to fix permanently.

Fair enough - if I were really paranoid I would've pushed for the
comprehensive solution rather than just fixing the problems I could
find - but I figured I'd mention it in the CR in case anyone had some
good ideas/discussion.

> If you want to fix it, maybe introduce a set of warnings to lit.py? I
> dunno... I'm not sure what the correct fix is, but I don't think it is to
> make the user-facing Clang driver produce more errors or support more flags.
> This is a problem for developers, and we should shield any solutions to it
> from the user-facing interfaces.

The immediate issue is for developers, but I just thought this might
be also biting users without them realizing it (as we didn't) when
passing flags that were being ignored just because it was a typo.
Perhaps I'm mistaken.

- David