Re: gnulib:test-stdalign failing on powerpc-darwin8, gcc-4.0

[David Fang]

Some more details:

On powerpc-darwin8, where the test fails, gcc is:
Using built-in specs.
Target: powerpc-apple-darwin8
Configured with: /var/tmp/gcc/gcc-5370~2/src/configure --disable-checking 
-enable-werror --prefix=/usr --mandir=/share/man 
--enable-languages=c,objc,c++,obj-c++ 
--program-transform-name=/^[cg][^.-]*$/s/$/-4.0/ 
--with-gxx-include-dir=/include/c++/4.0.0 --with-slibdir=/usr/lib 
--build=powerpc-apple-darwin8 --host=powerpc-apple-darwin8 
--target=powerpc-apple-darwin8
Thread model: posix
gcc version 4.0.1 (Apple Computer, Inc. build 5370)
(test fails with and without your patch to stdalign.in.h)


On powerpc-darwin9, where the test passes, gcc is:
sing built-in specs.
Target: powerpc-apple-darwin9
Configured with: /var/tmp/gcc/gcc-5493~1/src/configure --disable-checking 
-enable-werror --prefix=/usr --mandir=/share/man 
--enable-languages=c,objc,c++,obj-c++ 
--program-transform-name=/^[cg][^.-]*$/s/$/-4.0/ 
--with-gxx-include-dir=/include/c++/4.0.0 --with-slibdir=/usr/lib 
--build=i686-apple-darwin9 --program-prefix= --host=powerpc-apple-darwin9 
--target=powerpc-apple-darwin9
Thread model: posix
gcc version 4.0.1 (Apple Inc. build 5493)

Slightly different build numbers between darwin8 and 9.

I also tested Apple's gcc-4.2 on powerpc-darwin9:
Using built-in specs.
Target: powerpc-apple-darwin9
Configured with: /var/tmp/gcc_42/gcc_42-5577~1/src/configure 
--disable-checking --prefix=/usr --mandir=/usr/share/man 
--enable-languages=c,objc,c++,obj-c++ 
--program-transform-name=/^[cg][^.-]*$/s/$/-4.2/ --with-slibdir=/usr/lib 
--build=i686-apple-darwin9 --with-gxx-include-dir=/usr/include/c++/4.0.0 
--program-prefix= --host=powerpc-apple-darwin9 
--target=powerpc-apple-darwin9
Thread model: posix
gcc version 4.2.1 (Apple Inc. build 5577)

This too passes.

I'm pretty certain that powerpc-darwin8 and darwin9 have the same ABI 
struct-layout and alignment rules.
Something must be going wrong with the stdalign macros, specific to 
darwin8's compiler.

At this point, I have to consider trying other compilers on darwin8,
like one that supports and includes stdalign.h with built-in align macros.

Fang

> Hi Paul,
> 	I tried applying the patch (just the stdalign.in.h bit) to the 
> powerpc-darwin8 build, but I still see (when changing the assert to printf):
>
> static_char_alignas 5
> static_char_Alignas 4
> static_short_alignas 2
> static_short_Alignas 0
> static_int_alignas 4
> static_int_Alignas 0
> static_long_alignas 4
> static_long_Alignas 0
> static_int64_t_alignas 0
> static_int64_t_Alignas 0
> static_float_alignas 4
> static_float_Alignas 0
> static_double_alignas 0
> static_double_Alignas 0
> static_struct1_alignas 5
> static_struct1_Alignas 4
> static_struct2_alignas 0
> static_struct2_Alignas 6
> static_struct3_alignas 0
> static_struct3_Alignas 4
> static_struct4_alignas 0
> static_struct4_Alignas 4
>
> I also happened to test GNTLS-3.3.9 (containing gnulib) on powerpc-darwin9, 
> also using apple-gcc-4.0, before patching, and it passed all tests.
>
> I'm not entirely sure what is different between them.
> If you have any suggestions, I can try them tonight.
>
> Fang
>
> >  David Fang wrote:
> > >   Does this mean that it falls back to the distributed copy in
> > >   gl/stdalign.h?
> >
> >  Yes, actually, it's basically a copy of gl/stdalign.in.h.
> >
> > >   I don't see any evidence of the replacement stdalign.h being tested in
> > >   config.log.
> >
> >  The replacement is not tested, because its runtime behavior can't easily
> >  be tested with cross-compilation.
> >
> >  Do you observe the same problem with Apple's GCC 4.2?
> >
> >  Do the attached patches work around the problem for you?  The second one's
> >  for HP compilers and you can ignore it if you like.

--
David Fang
http://www.csl.cornell.edu/~fang/