Re: GNU SASL test failures

[Simon Josefsson]

address@hidden (Ludovic Courtès) writes:

> Hi,
>
> Simon Josefsson <address@hidden> writes:
>
>> address@hidden (Ludovic Courtès) writes:
>>
>>> Hi Simon!
>>>
>>> Simon Josefsson <address@hidden> writes:
>>>
>>>> I just realized that the hydra build is using GNU GSS from the build
>>>> system, not the latest from git.  Do you know which version of GNU GSS
>>>> is installed on that system?
>>>
>>> GSS 1.0.0.  (You can see it by going to a build page, e.g.,
>>> <http://hydra.nixos.org/build/413115>, then clicking on the
>>> “Information” tab, “build-time dependencies”, “list”, and searching for
>>> ‘gss’.)
>>
>> Ah -- I was looking for gss/shishi under 'Build inputs', what is the
>> difference between a build dependency and a build input?
>
> “Build Inputs” are dependencies managed by Hydra, which is anything
> built from VCS.
>
> “Build dependencies” include all the dependencies, notably those dragged
> from Nixpkgs (the compiler, C library, GSS in this case, etc.)
>
> Subtle, I have to admit.  ;-)

Ah, I see.

>> Should be fixed in git now.  Could you try to enable Shishi in GSS again
>> (or maybe it happens automatically)?
>
> In that case GSS was taken from Nixpkgs, i.e., from a release tarball.
> So no that won’t happen automatically.
>
> We could add GSS to Hydra, and make it a “build input” of SASL, in which
> case things would happen automatically.
>
> In the meantime, I’ve worked around it in Nixpkgs by doing
> “make check LDFLAGS=-lshishi”.  Hydra will pick it up and it will
> eventually propagate to the SASL builds.

Thanks.  I'll release GSS 1.0.1 soon with the fix, but it would be great
if GSS (and Shishi) could be added to Hydra as well.  Then I know for
sure whether the fix I applied solves the problem or not.

/Simon