Continuum and build errors handling

[Julien Stern <ju...@cryptolog.com>]

Hi list,

We have the following issue with continuum:

On some occasion, a build fails because of an "error".
For instance, the CVS server could be temporarily down, or
the repository, or one disk space could be full,
or there could be an OutOfMemoryError, or the time limit could
be reached because some heavy task is running on the box at the
same time (backups...), etc...

In that case, the corresponding project is never rebuilt
(well, until the next commit or dependency change...)
and the rebuild needs to be triggered manually.

While I agree it is probably not a good idea to systematically try to
rebuild a project that encountered an error, don't you think it could
be a worthwhile addition to continuum to allow for a small number of
retries in such a case?

More precisely, when a project cannot be built because of an "error", it
could be put in the queue again for a few (say 3, 4) retries.
That would allow for automatic handling of transient errors without
putting too much stress of the server by always rebuilding projects vowed
to failure.

Regards,

--
Julien