Re: [bug-mes] ARM mes on wip-arm - Other test failures

[Jan Nieuwenhuizen]

Danny Milosavljevic writes:

>> FAIL: lib/tests/scaffold/7l-struct-any-size-array-simple.c
>
> char is unsigned on ARM by default.  Can be changed by gcc option 
> "-fsigned-char".

Ah!  I wondered about that and chose the naughty option "never ever seen
unsigned char", let's assume that's only a philosophical option.  Hmm.

> The C standard is intentionally vague there, so it probably should be
> made to work fine with unsigned-by-default char, too, instead.

Yeah, or disable for arm.  Hmm, how is tcc compiled?  A simple grep does
not reveal forcing char to signed for arm.

>> FAIL: lib/tests/scaffold/7r-sign-extend.c [a char with value -129 ? Is that 
>> valid C?  In any case, more fundamental things are broken]
>> FAIL: lib/tests/setjmp/80-setjmp.c [not implemented for __arm__]
>> FAIL: lib/tests/scaffold/85-sizeof.c [has no case for __arm__]
>> FAIL: lib/tests/dirent/90-readdir.c [can open nonexistent directory]
>
> It's opening a regular file as a directory and succeeding (?).

Hmm...fishy.  So probably the mes c library needs to learn more about
what the kernel returns?

>> FAIL: lib/tests/io/90-stat.c [no case for __arm__]
>> FAIL: lib/tests/mes/90-dtoab.c [problem with negative floating point numbers]
>> FAIL: lib/tests/signal/90-signal.c [signal handler doesn't get called]
>> XFAIL: lib/tests/stdio/90-sprintf.c

Thanks!
janneke

-- 
Jan Nieuwenhuizen <address@hidden> | GNU LilyPond http://lilypond.org
Freelance IT http://JoyofSource.com | Avatar® http://AvatarAcademy.com