r185544 - Fix PR16454: Don't #include altivec.h when preprocessing assembly.

[Richard Smith]

On Wed, Jul 3, 2013 at 11:02 AM, Bill Schmidt
<wschmidt at linux.vnet.ibm.com> wrote:
> On Wed, 2013-07-03 at 10:50 -0700, Richard Smith wrote:
>> On Wed, Jul 3, 2013 at 10:31 AM, Bill Schmidt
>> <wschmidt at linux.vnet.ibm.com> wrote:
>> > On Wed, 2013-07-03 at 10:19 -0700, Eli Friedman wrote:
>> >> On Wed, Jul 3, 2013 at 8:36 AM, Bill Schmidt
>> >> <wschmidt at linux.vnet.ibm.com> wrote:
>> >>         Author: wschmidt
>> >>         Date: Wed Jul  3 10:36:02 2013
>> >>         New Revision: 185544
>> >>
>> >>         URL: http://llvm.org/viewvc/llvm-project?rev=185544&view=rev
>> >>         Log:
>> >>         Fix PR16454: Don't #include altivec.h when preprocessing
>> >>         assembly.
>> >>
>> >>         When the -maltivec flag is present, altivec.h is auto-included
>> >>         for the
>> >>         compilation.  This is not appropriate when the job action is
>> >>         to
>> >>         preprocess a file containing assembly code.  So don't do that.
>> >>
>> >>         I was unable to convert the test in the bug report into a
>> >>         regression
>> >>         test.  The original symptom was exposed with:
>> >>
>> >>           % touch x.S
>> >>           % ./bin/clang -target powerpc64-unknown-linux-gnu -maltivec
>> >>         -S -o - x.S
>> >>
>> >>         I tried this test (and numerous variants) on a PPC64 system:
>> >>
>> >>         ----------------------------------------------------------------------------
>> >>         // RUN: touch %t
>> >>         // RUN: %clang -maltivec -S %t -o - | FileCheck %s
>> >>
>> >>         // Verify that assembling an empty file does not auto-include
>> >>         altivec.h.
>> >>
>> >>         // CHECK-NOT: static vector
>> >>         ----------------------------------------------------------------------------
>> >>
>> >>         However, this test passes for some reason even on a clang
>> >>         built
>> >>         without the fix.  I'd be happy to add a test case but at this
>> >>         point
>> >>         I'm not able to figure one out, and I don't want to hold up
>> >>         the patch
>> >>         unnecessarily.  Please let me know if you have ideas.
>> >>
>> >>
>> >> Umm, why are you committing a patch for an issue you can't reproduce?
>> >
>> > Sorry if I was unclear.  I can reproduce the issue by hand.  I can't
>> > find a way to automate the process successfully using FileCheck.  I
>> > don't know why it doesn't work as a test case when it works from the
>> > command line.  No doubt there is some small thing about the automatic
>> > testing process that I don't understand.
>> >
>> > I had posted this potential solution a couple of days ago in the bug,
>> > and asked for assistance with the test case.  After no response, I
>> > decided to go ahead with the patch and add a test case later if I can
>> > get some help in understanding why it doesn't work.
>>
>> So... comments in bugs don't get attention from people who aren't
>> CC'd; even people subscribed to the llvmbugs@ list only see when bugs
>> are created and closed, not comments added to them. In future, please
>> send patches for which you want review or comments to cfe-commits@
>> (even if they're incomplete); you're vastly more likely to get a
>> response there. Also, as a general rule, please don't interpret a lack
>> of response as a signal to go ahead, the expectation within the
>> community is that you will ping a patch if you don't get an answer to
>> your initial mail (around once a week is typical).
>
> Sorry if this wasn't quite kosher.  My intent was to request broader
> help with the test case with this patch note.  I didn't want to hold up
> the original filer of the bug over this as I will be out of town for a
> number of days after today.  But yes, I understand the process and will
> be more strict about it in future.
>
> Do you have a preference for the location of this test case
> (test/Driver, test/CodeGen, etc.)?

test/Driver seems appropriate. Thanks for adding the test case!