This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: XFAIL for GCC compat testsuite


Janis Johnson <janis187@us.ibm.com> writes:

> On Thu, Nov 13, 2003 at 12:20:08PM +0100, Eric Botcazou wrote:
>> Hi Janis,
>> 
>> I've added the possibility to XFAIL your compat testcases on compilation in 
>> my tree (for the SPARC port), using the new dg-xfail-if infrastructure.
>> 
>> Tested on i586-redhat-linux-gnu, sparc64-sun-solaris2.9 and 
>> sparc-sun-solaris2.8. Ok for mainline?
>> 
>> 
>> 2003-11-13  Eric Botcazou  <ebotcazou@libertysurf.fr>
>> 
>> 	* lib/compat.exp (compat-obj): New xfaildata parameter.
>> 	Use it to set compiler_conditional_xfail_data before compiling.
>> 	(compat-get-options): Handle dg-xfail-if.
>> 	(compat-execute): Retrieve XFAIL data and pass them to compat-obj.
>> 	* gcc.dg/compat/vector-1_x.c: XFAIL on SPARC.
>> 	* gcc.dg/compat/vector-1_y.c: Likewise.
>> 	* gcc.dg/compat/vector-2_x.c: Likewise.
>> 	* gcc.dg/compat/vector-2_y.c: Likewise.

The patch is OK.

> As the author of the original code this looks good to me, but I'm not
> a maintainer and can't approve it.

You ought to be considered official maintainer of the compat
testsuite, IMHO.

> To whichever maintainer looks at this, please also take a look at the
> fix in http://gcc.gnu.org/ml/gcc-patches/2003-10/msg01114.html from
> David Ung; there was a discussion just before this about how no test
> case change could be considered obvious, otherwise I would have checked
> it in.

Bringing vector-check.h into consistency with vector-setup.h is
clearly the right thing.  I would not presume to say which of the two
should change, but since you agree that David's patch is correct, go
for it.

As for the "no obvious change to the test suite" thing...  The
concerns raised in that discussion mostly have to do with how it's
easy to confuse an automated regression robot into thinking something
has regressed, or un-regressed, when it hasn't, by making changes to
the test suite.  Further, an awful lot of the GCC test suite is
meaningful only insofar as it used to provoke some specific bug, which
has long since been fixed and forgotten.  Those parts of the test
suite cannot be changed without very careful consideration, lest the
salient part of the test case be lost.  However, your compat suite is
not like that; it has a design, we know what it's trying to test, and
when there's a bug in it (which this is) which is clearly orthogonal
to the thing being tested, a fix can be obvious.

zw


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]