Re: coreutils-5.1.0 build feedback

[Jim Meyering]

Thanks for all the testing!

"Nelson H. F. Beebe" <address@hidden> wrote:
> On DEC Alpha GNU/Linux, the coreutils-5.1.0 test suite had one failure:
>
> FAIL: fold

I'll investigate.

> On MacOS X, there was one failure:
>
> FAIL: basic
> ./deref: skipping this test; your system doesn't support changing
> the owner or group of a symbolic link.

Fixing that may be as simple as appending ` 77' to this line
of tests/chgrp/deref:

  (exit 77); exit

> On Sun Solaris 2.7 with CC=c89, there was one test failure:
>
> out exp differ: char 37, line 1
> split: 4294967296: invalid number of lines
> Try `split --help' for more information.
> split: 4294967296: invalid number of bytes
> Try `split --help' for more information.
> 1c1
> < split: line count option -9999999999... is too large
> ---
>> split: line count option -99999999999999999991... is too large
> FAIL: split-fail
>
> This happened on Solaris 2.8 and 2.9 as well.

Sounds like uintmax_t resolves to a 32-bit type with c89.
If you could direct me to such a system, I'll investigate it.

> On GNU/Linux on Intel IA-32 with Intel icc, there was a compilation
> failure:
>
> icc -DHAVE_CONFIG_H -I. -I. -I..  -I.. -I. -I../lib -I../lib
> -I/usr/local/include  -g -c stat.c
> /usr/include/bits/statfs.h(27): error: identifier "__SWORD_TYPE" is undefined
>       __SWORD_TYPE f_type;
>       ^

Do I have access to such a system?