[PATCH 6/6 v4]: Merge from Stack Branch - Test Cases

[H.J. Lu]

Hi Janis,

On Thu, Jun 5, 2008 at 3:56 PM, Janis Johnson <janis187@us.ibm.com> wrote:
> On Thu, 2008-06-05 at 16:03 +0800, Ye, Joey wrote:
>> Updated version.
>>
>> 2008-06-05  H.J. Lu  <hongjiu.lu@intel.com>
>>           Joey Ye  <joey.ye@intel.com>
>>
>>       * gcc.target/i386/20060512-1.c: Add
>>         -mpreferred-stack-boundary=4.
>>       (main): Move "popl" after check.
>>       * gcc.target/i386/20060512-3.c: Likewise.
> [lots more]
>
> When a test has been copied from another part of the testsuite, add a
> comment giving the path of the original.

All stack alignment run-time testscases are designed to
test stack alignment. Some of them may be modified
or combined based on the existing tests. But they are
different tests. The purpose of those tests is to check
the alignment of local variable in different conditions
at run-time.

>
> Avoid using C library functions and their headers whenever possible.
> For example, instead of including <string.h> and calling strncpy, call
> __builtin_strncpy.
>

Joey has made the change.

> Add comments to tests in which the purpose isn't clear, as with
>  gcc.target/i386/stackalign/asm-1.c
>  gcc.target/i386/stackalign/return-2.c
> They compile with no additional checks; did they cause ICEs in earlier
> versions on the branch?
>
> Torture option handling was cleaned up recently, and it should be
> possible to make gcc.dg/torture/stackalign/stackalign.exp much
> simpler by combining the existing C_TORTURE_OPTIONS with your
> sets of additional options.

I will take a look when I find time.

> Just FYI (I don't mind if it stays the way it is), this:
>
>  /* { dg-do run } */
>  /* { dg-require-effective-target ilp32 } */
>  /* { dg-require-effective-target dfp } */
>
> is the same as this:
>
>  /* { dg-do run { target { ilp32 && dfp } } } */
>
> or this:
>
>  /* { dg-do run } */
>  /* { dg-require-effective-target { ilp32 && dfp } } */
>
> Unless there are other objections you can add these tests when the
> functionality has been committed.
>

Here is the updated patch against trunk. We will wait until
the rest of patches are reviewed and approved.

Thanks.

-- 
H.J.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: test.txt
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20080606/71b46726/attachment.txt>