[cfe-dev] c11.c/cxx11.cpp tests are failing for Hexagon

Daniel Dunbar daniel at zuster.org
Thu Apr 18 10:55:02 PDT 2013

Hi Jyotsna,

You are right this is poor behavior in the tests (which shouldn't ever
depend on the system headers).

I believe the c11.c test should be fixed here:

Is that not the case?

 - Daniel

On Wed, Apr 17, 2013 at 12:19 PM, Richard Smith <richard at metafoo.co.uk>wrote:

> On Wed, Apr 17, 2013 at 10:39 AM, Jyotsna Verma <jverma at codeaurora.org>wrote:
>> Hi Daniel/Richard,
>> I'm seeing c11.c and cxx11.cpp tests failing on Hexagon with these two
>> changes - r179427 and r179419.
>> After your changes, we include stdint.h file (from lib/Headers) in both
>> these tests. However,  if  __STDC_HOSTED__ macro is defined,
>> lib/Headers/stdint.h includes system stdint.h file which includes
>> features.h
>> -> gnu/stubs-32.h. Since, we don't have stubs-32.h header file available
>> for
>> Hexagon, it causes the tests to fail.
>> Please help me understand why the tests were modified to include stdint.h
>> file and how to fix them for Hexagon.
> These tests are testing the behavior of our <stdint.h> implementation. It
> would certainly make sense to make them avoid including a system-provided
> <stdint.h>. That will probably require somewhat different fixes for each
> test:
> I think (perhaps Daniel can confirm) that the c11.c test just wants to
> test the contents of our own <stdint.h>, so modifying the test's RUN: line
> so that the system-provided <stdint.h> is not on the search path would
> suffice.
> For the cxx11.c test, what we're checking is that our builtin <stdint.h>
> correctly works around http://sourceware.org/PR15366 -- though a better
> (non-system-dependent) test here would be to craft a "system header"
> <stdint.h> with the C99-specified behavior and put that in the system
> headers include path for the test.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20130418/f371d47b/attachment.html>

More information about the cfe-dev mailing list