On Mon, Nov 11 2013, David Bremner <david@tethera.net> wrote:
> David Bremner <david@tethera.net> writes:
>
>> Daniel Kahn Gillmor <dkg@fifthhorseman.net> writes:
>>
>>>
>>> Please don't introduce this cruft into the notmuch codebase.  It should
>>> be fixed in gmime, not worked-around notmuch.
>>>
>>> I've just uploaded gmime 2.6.19-2 to unstable to address this issue.
>>>
>>
>> Hi Daniel;
>>
>> Thanks a lot for that. 
>>
>> What I (still) wonder about is all the people not running Debian, in the
>> interval between the release of notmuch 0.17 and the next upstream
>> release of gmime (and propagation to various distros).  Even on Debian,
>> building on testing and backports complicates things a bit.
>
> An alternative approach would be to fix 
>
>  NOTMUCH_SKIP_TESTS="reply reply-to-sender multipart.37 from-guessing" ./notmuch-test
>
> to return exist status 0, and ideally to print the correct number of
> tests skipped.
>
> I think the former just requires patching line 85 of
> aggregate-results.sh.  I don't see an easy way to do the latter.
>
> This is arguably less crufty, or at least uses existing cruft. On the
> other hand, it doesn't distinguish between a test printing an extra
> space, and complete garbage.

I agree with Daniel that the original series adds a whole lot of
undesirable cruft to the test suite to get around one transient issue in
one of the notmuch dependencies.  That doesn't seem like the best thing
to do imho.  I would vote for this option instead, which doesn't modify
the test suite and just skips the affected tests until the new gmime
versions falls through everywhere.

jamie.