On 6/20/2016 5:39 PM, Jeff Law wrote:
> On 06/20/2016 09:26 AM, Tim Prince wrote:
>> I note that a few of the tests in gcc testsuite are marked to show stabs
>> tests as unsupported:
>>
>> pr35154.c:/* { dg-skip-if "No stabs" { mmix-*-* alpha*-*-* hppa*64*-*-*
>> ia64-*-*
>> *-*-vxworks* } { "*" } { "" } } */
>> pr35154.c:/* { dg-skip-if "stabs only" { *-*-* } { "*" } { "-gstabs"
>> } } */
>>
>> Why are there hundreds of tests which aren't so marked and fail on
>> targets which lack stabs support? Are we expected to edit out the
>> failure lines which include stabs before sending to gcc-testresults?
> What targets? They may simply be targets that aren't getting a lot of
> testing attention and associated maintenance. And stabs certainly
> don't get much attention.
>
> In an ideal world stabs would just die :-)
>
>
>
>>
>> What about coff tests?
> Not sure what you're asking here.
>
> jeff
My impression is that coff isn't supported on Windows either, yet there
are a few tests which always fail.
Another impression I have is that several Windows targets get a fair
amount of usage, with both good and not-so-good binaries available, but
relatively little posting on testresults. I suppose that may be partly
on account of the excessive reporting of FAIL for unsupported features,
so no one knows which FAILs indicate work to be done and all are ignored.
If anyone cares about binaries which appear to be distributed in
violation of licensing, or simply without adequate support, perhaps
better testsuite support of those extremely useful ones which try to
comply might be beneficial.
cygwin just added support for the long double math functions needed to
pass certain cases.
I still haven't seen any way to make testsuite run parallel correctly on
Windows, so it's about a 40 hour slog unless distributed across several
independent boxes.
Thanks.
--
Tim Prince
