[cfe-commits] r164677 -

[Sean Silva]

>> Alternatively (and slightly more generally) how about teaching -verify to
>> fail if it doesn't find any expected-* comments to check (like FileCheck
>> does)?
>
> That wouldn't have helped in this case though, would it? there's no
> expected- comment in this file.

Wut? I think that what Richard was proposing elegantly addresses this
case. Basically, it fails when it doesn't see an expected-* comment.
Since stdin is empty, then there would be no expected-* comment, so
the test would fail.

Richard, how do you suggest the diagnostic should be like? I fear that
an error like "didn't see an expected-* comment" wouldn't be that
helpful in diagnosing the issue at hand (missing input file,
defaulting to stdin), even though it does accurately reflect the
situation. As a hack, a note could be added, but that doesn't seem
very in line with Clang style.

Maybe it is worthwhile to require an explicit input file AND fail if
no expected-* comments are found?

> Sounds resonable. Wanna give it a shot? :-)

Yeah, once we settle on a way to do this I'll prepare a patch (unless
someone beats me to it).

--Sean Silva

On Thu, Sep 27, 2012 at 10:26 PM, Nico Weber <thakis at chromium.org> wrote:
> On Fri, Sep 28, 2012 at 10:24 AM, Richard Smith <richard at metafoo.co.uk> wrote:
>> On Thu, Sep 27, 2012 at 6:05 PM, Nico Weber <thakis at chromium.org> wrote:
>>>
>>> On Fri, Sep 28, 2012 at 1:31 AM, Sean Silva <silvas at purdue.edu> wrote:
>>> >> I couldn't think of a great way. If no filename argument is present,
>>> >> -cc1 -verify reads its input from stdin, which is a valid usecase. I
>>> >> think lit closes stdin when it forks, which is why -cc1 doesn't just
>>> >> wait for eof forever in this case.
>>> >
>>> > It is a valid use case, but how about just requiring that it be
>>> > explicit (at least for -verify)? Forcing it to be explicit would
>>> > preserve the use case but make the error case impossible.
>>>
>>> Sounds resonable. Wanna give it a shot? :-)
>>
>>
>> Alternatively (and slightly more generally) how about teaching -verify to
>> fail if it doesn't find any expected-* comments to check (like FileCheck
>> does)?
>
> That wouldn't have helped in this case though, would it? there's no
> expected- comment in this file.
>
>>
>>>
>>> >
>>> > -- Sean Silva
>>> >
>>> > On Thu, Sep 27, 2012 at 6:22 AM, David Blaikie <dblaikie at gmail.com>
>>> > wrote:
>>> >>  /cfe/trunk/test/Sema/no-format-y2k-turnsoff-format.c
>>> >> MIME-Version: 1.0
>>> >> Content-Type: text/plain; charset="utf-8"
>>> >> Content-Transfer-Encoding: 7bit
>>> >>
>>> >> If clang was going to read from stdin, wouldn't it need a -x to tell it
>>> >> what kind of input it is? Even if it has a default, an empty file is
>>> >> not a valid TU, at least in C++, though I guess if it has a default it
>>> >> would be C and an empty file is probably valid C. Perhaps a warning
>>> >> there at least could be handy
>>> >> From: Nico Weber
>>> >> Sent: 9/27/2012 7:43 AM
>>> >> To: Sean Silva
>>> >> Cc: Nico Weber; cfe-commits at cs.uiuc.edu
>>> >> Subject: Re: [cfe-commits] r164677
>>> >> - /cfe/trunk/test/Sema/no-format-y2k-turnsoff-format.c
>>> >> On Thu, Sep 27, 2012 at 2:46 AM, Sean Silva <silvas at purdue.edu> wrote:
>>> >>> Is there some way to ensure that this doesn't ever happen again? Like
>>> >>> having -verify warn/fail if there is no input? That would allow fixing
>>> >>> all of these cases in a single fell swoop and ensure that it never
>>> >>> happens again.
>>> >>
>>> >> I couldn't think of a great way. If no filename argument is present,
>>> >> -cc1 -verify reads its input from stdin, which is a valid usecase. I
>>> >> think lit closes stdin when it forks, which is why -cc1 doesn't just
>>> >> wait for eof forever in this case.
>>> >>
>>> >> I suppose lit could check if clang_cc1 is the first command on the run
>>> >> line and then add a flag that tells cc1 to error out if it's reading
>>> >> from stdin, but seems brittle.
>>> >>
>>> >> Do you have a good suggestion?
>>> >>
>>> >>>
>>> >>> -- Sean Silva
>>> >>>
>>> >>> On Wed, Sep 26, 2012 at 5:02 AM, Nico Weber <nicolasweber at gmx.de>
>>> >>> wrote:
>>> >>>> Author: nico
>>> >>>> Date: Wed Sep 26 04:02:07 2012
>>> >>>> New Revision: 164677
>>> >>>>
>>> >>>> URL: http://llvm.org/viewvc/llvm-project?rev=164677&view=rev
>>> >>>> Log:
>>> >>>> Make this test actually test something
>>> >>>>
>>> >>>> Modified:
>>> >>>>     cfe/trunk/test/Sema/no-format-y2k-turnsoff-format.c
>>> >>>>
>>> >>>> Modified: cfe/trunk/test/Sema/no-format-y2k-turnsoff-format.c
>>> >>>> URL:
>>> >>>> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Sema/no-format-y2k-turnsoff-format.c?rev=164677&r1=164676&r2=164677&view=diff
>>> >>>>
>>> >>>> ==============================================================================
>>> >>>> --- cfe/trunk/test/Sema/no-format-y2k-turnsoff-format.c (original)
>>> >>>> +++ cfe/trunk/test/Sema/no-format-y2k-turnsoff-format.c Wed Sep 26
>>> >>>> 04:02:07 2012
>>> >>>> @@ -1,4 +1,4 @@
>>> >>>> -// RUN: %clang_cc1 -verify -fsyntax-only -Wformat -Wno-format-y2k
>>> >>>> +// RUN: %clang_cc1 -verify -fsyntax-only -Wformat -Wno-format-y2k %s
>>> >>>>  // rdar://9504680
>>> >>>>
>>> >>>>  void foo(const char *, ...) __attribute__((__format__ (__printf__,
>>> >>>> 1, 2)));
>>> >>>>
>>> >>>>
>>> >>>> _______________________________________________
>>> >>>> cfe-commits mailing list
>>> >>>> cfe-commits at cs.uiuc.edu
>>> >>>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>> >>> _______________________________________________
>>> >>> cfe-commits mailing list
>>> >>> cfe-commits at cs.uiuc.edu
>>> >>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>> >> _______________________________________________
>>> >> cfe-commits mailing list
>>> >> cfe-commits at cs.uiuc.edu
>>> >> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>> _______________________________________________
>>> cfe-commits mailing list
>>> cfe-commits at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>
>>