Re: Requirements for snapshot (was Re: gcc-ss-20001002 is now available

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

References: <20001002205559.A20847@disaster.jaj.com> <20001002175356.J19698@namaste.lokigames-lan.com>
Reply-to: rittle@rsch.comm.mot.com
BCC: rittle@latour.rsch.comm.mot.com
--text follows this line--
In article <20001002175356.J19698@namaste.lokigames-lan.com>,
Michael Vance <briareos@lokigames.com> wrote:

>> I was under the impression that, while snapshots weren't guaranteed to be
>> stable, they would at least bootstrap.

This is a mistaken impression.  Contributors are suppose to ensure
that any changes they check-in to the master CVS repository for gcc
bootstrap (or simpler rebuild process, depending upon which files are
changed) and 'make check' on at least one target related to the change
(i.e. if you change the configuration files for Linux/IA32, you can't
bootstrap Linux/PPC and claim that you did your job).  But, sometimes,
other targets do break unexpectantly.  On rare occasion, things do get
checked in that completely break the tree for most people since
(actual problems from my memory:) there was a mistake in testing,
there was a mistake in committing the change or there was an unknown
dependency on a quirk of the submitter's development environment.

> So was I. I actually thought they were regression-tested, also.

This is also a mistaken impression.  It is expected that anyone
concerned about gcc's future quality will build and regression test
the snapshots and report their results to the gcc-testresults list.

Regards,
Loren