Re: xassert in dispextern.h

[David Kastrup <dak@gnu.org>]

Jason Rumney <jasonr@gnu.org> writes:

> David Kastrup <dak@gnu.org> writes:
>
>> That's completely backwards.
>
> What's completely backwards is to turn off a feature that helps us
> find bugs so that people can treat the CVS HEAD as a stable release
> of Emacs.

It does not help us find bugs.  That's the problem.  It's been on for
a month now, and the only bugs that were triggered by it were of the
sort that can much easier be found, reproduced and described when the
code does not abort.  I have wasted about 4 days of debugging on
something that turned out not be really a bug (in the sense that it
could lead to data destruction) but a quirk; and if the assertion had
not been turned on, both the severity and kind of this quirk would
have been _much_ better accessible for estimation.

GLYPH_DEBUG is not on by default either.  It is good to have the
assertions for tracking down a particular problem.  But I have seen no
evidence whatsoever that switching the default has helped us finding
even a single bug.

In contrast, it has kept me for several days from fixing real bugs and
making a release with other software.

And it also means that we have to tell people "don't use CVS Emacs, it
is slow and will crash".  And that means that all those people won't
help in finding _real-life_ bugs occuring in serious daily
application.  We don't have a formal beta program.  Getting this
feedback is necessary for ensuring a good quality release.

Feel free to volunteer any differing information if you have it
available.  If you know of any problem in the last 4 weeks that has
been discovered and fixed due to the changed default, I'd be glad to
hear of it.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum