Subject: Re: speed results
From: Jesper Skov (jskov@cambridge.redhat.com)
Date: Sat Apr 14 2001 - 11:00:59 CDT
>>>>> "Martin" == Martin Sevior <msevior@mccubbin.ph.unimelb.edu.au> writes:
Martin> Then toss in some undos/redos. If you really want to break Abi
Martin> put in some section breaks with column changes then undo/redo
Martin> them.
That actually works :) But paste section breaks, and you get a
segv. I'm just looking at bug 800. Hope to fix it today.
But a torture test would be really useful: not just a specific test,
but one that has a list of all operators and randomly executes
these. It should keep a log so it can be replayed if it does trigger
a crash (or a scream, to keep with the terminology :)
Whenever a crash is found, add a bug, and the particular log as a
regression test. Obviously, first make sure it's not a duplicate of an
existing known problem. But this would be a neat way to test out
corner cases - I'm sure we have at least a handful of unknown segvs
hiding in the code which only get triggered by daft operations that
only someone like me would do - and I don't have the time to play user :)
Jesper
This archive was generated by hypermail 2b25 : Sat Apr 14 2001 - 11:02:29 CDT