Re: Small update to reversed_comparison_code

["Joseph S. Myers" <jsm28 at cam dot ac dot uk>]

On Tue, 13 Mar 2001, Mark Mitchell wrote:

>   In fact, one of our SC members has suggested automatically removing
> patches that fail Geoff's regression checker, and others have gone so
> far as to suggest requiring an automated cross-platform bootstrap
> *before* check-in.

As a start, automated checkers that point out regressions on the primary
platforms are needed.  Perhaps CodeSourcery's autobuilder could change
from a daily one to a round-the-clock one, that points out regressions to
gcc-regression and to the apparently responsible people, as with Geoff's
one?  This would at least double the number of platforms checked.

Doing any of this properly probably needs three or more machines per
platform (though if fast enough the machines could do multiple tasks in
parallel):

* One to build round the clock and detect problems.  (In principle, one
for both mainline and release branch.)

* One to do a binary search on individual patches when the above when
detects problems, to narrow down when problems appeared.  (In principle,
one for both mainline and release branch.)

* One to which all developers have access to debug problems with their
patches.  Just knowing a patch causes a problem may not help otherwise.

Preferably, one should also every so often run past snapshots and releases
against the current mainline testsuite, to detect regressions in tests
added after the regression appeared.  This would need to do a binary
search on any regressions found to be of much use.

-- 
Joseph S. Myers
jsm28@cam.ac.uk