guile testsuite failure (gnutls 3.0.1 and later) and armel and mipsel

[Simon Josefsson]

Nikos Mavrogiannopoulos <nmav at gnutls.org> writes:

> On 10/24/2011 03:04 PM, Simon Josefsson wrote:
>> Andreas Metzler <ametzler at downhill.at.eu.org> writes:
>> 
>>> On 2011-10-19 Andreas Metzler <ametzler at downhill.at.eu.org> wrote:
>>>> On 2011-10-18 Ludovic Courtès <ludo at gnu.org> wrote:
>>> [...]
>>>>> Andreas: Could you bisect it yourself, between 3.0.0 and 3.0.1?
>>>
>>>> I probably won't have time before the weekend, but I will try then if
>>>> the issue is still open.
>>> [...]
>>>
>>> The commit which triggers the bug is this one:
>>> |-----------------------------
>>> | From 2385c7f999c12802b11859a34b89ff7662b1f4af Mon Sep 17 00:00:00 2001
>>> | From: Nikos Mavrogiannopoulos <nmav at gnutls.org>
>>> | Date: Sat, 13 Aug 2011 14:10:55 +0200
>>> | Subject: [PATCH] Added crywrap to the distributed programs.
>>> |-----------------------------
>>>
>>> Apart from adding crywrap it adds a lot of new gnulib modules, I guess 
>>> one of these causes the problem.
>> Which gnulib modules?  Some modules don't work well with shared
>> libraries, and must be in a separate gnulib directory only linked to by
>> the applications (and not the library).
>
> Modules like argp,dirname,dirent etc. They are visible at:
> http://git.savannah.gnu.org/gitweb/?p=gnutls.git;a=commitdiff;h=2385c7f999c12802b11859a34b89ff7662b1f4af
>
> if you check the additions in gl/Makefile.am. However they aren't used
> by gnutls.

None of the added modules looks suspicious, however maybe one of them
depends on some lower level header file that causes the issue.  However,
the problem Ludo' diagnosed does not seem related to any system header
file to me -- instead it seems much more related to guile system library
confusion (e.g., having multiple guile libraries installed).

Andreas, how did you identify that the patch above is responsible?  Were
you able to reproduce this on a x86/amd64 system?

/Simon