Re: Testresults for gawk 3.1.7

[Aharon Robbins]

Hi.

> From: Arthur Loiret <address@hidden>
> Newsgroups: gnu.utils.bug
> Subject: Testresults for gawk 3.1.7
> Date: Sat, 7 Nov 2009 17:15:02 +0100
> To: address@hidden
> 
> Hello,
> 
> gawk 3.1.7 has been uploaded to Debian experimental, but it seems
> there are failures in the testsuite:
> 
> i386 (lc_num1, mbfw1, mbprintf1)
> >>> test/_lc_num1
> 1000000 1000000
> >>> test/_mbfw1
> Site
>   Kungsbacka
>  l=C3=83=C2=A4nG=C3=83=C2=B6teborg
> Stenungsund
> >>> test/_mbprintf1
> AAA    |
> =C3=83=E2=80=A6=C3=83=C6=92=C3=83=E2=80=A0 |

This may be a locale issue, but on i386 Linux there should not be ANY
failures as that is currently what I use.  Does this version of Debian
use GLIBC?

> sparc (lc_num1, mbfw1, mbprintf1)

These are locale and and machine dependent; I don't think you need
to worry about them.

> armel/ARM EABI (lc_num1, mbfw1, mbprintf1)
> s390/IMB Z90 (lc_num1, mbfw1, mbprintf1)
> alpha/DEC Alpha (double1, double2, intformat, lc_num1, mbfw1, mbprintf1)

Same here.

> The full builds logs are available at
> http://experimental.ftbfs.de/build.php?arch=3D&pkg=3Dgawk
> More builds (mainly mips, amd64, powerpc, hppa, ia64) should come
> within the next days. Could you please tell me more about those
> failures?

Some of the tests are explicitly marked as machine dependent
because they are at the outer range of the precision of floating
point numbers, so it's OK if they don't give the same result.

Also, some libraries don't support the ' flag to %d to add
a thousands separator, and/or they don't handle multibyte
characters.

HTH,

Arnold