Re: TODO, made current

[Loren James Rittle <rittle at latour dot rsch dot comm dot mot dot com>]

In article <m3wulfbzfs.fsf@uniton.integrable-solutions.net>,
Gabriel Dos Reis<gdr@integrable-solutions.net> writes:

> Benjamin Kosnik <bkoz@redhat.com> writes:
> | People seem to be especially
> | unwilling to come up with MT test cases.

> I believe that part of the problem lies in the fact, usually, it is hard
> to reproduce those problems with short testcases.

Good point and I fully agree.  In fact, it seems that unless we were
to add some hooks in the library to allow an external agent to insert
controlled stall points during testing (or someone builds a
vgrind/purify-like automatic system for MT sync points instead of
memory allocation), no MT testcases will ever be 100% reproducible.  I
would disfavor any non-automated approach.

Regards,
Loren