Re: when emacs 22.1 release will ready?

[joakim@verona.se]

"Lennart Borgman (gmail)" <lennart.borgman@gmail.com> writes:

>> With a similar release procedure for Emacs, Emacs 22.1 had been
>> released in 2004, and 23.4 would be ready for release next month.
>
> What do they do to keep control over the quality? I guess they are
> using a lot of unit testing for the kernel, or?
>
> Is that something that could be done for Emacs? I would propose that
> this should be done, in a proper manner. The way I would like to see
> it done is to add unit tests when relevant. There are a few points to
> watch out for:
>
> - Any change.
> - Any bug reports.
> - Any code that is hard to understand.
>
> A problem is of course that Emacs is screen oriented and that it is
> much harder to have an automatic testing framework for that. I do not
> know how people do that now, but semi-automatic result testing should
> be possible in cases for fully automatic testing is not possible.

I wrote some simple unit-test for CEDET
(http://www.emacswiki.org/cgi-bin/wiki/CollectionOfEmacsDevelopmentEnvironmentTools)

Having done so, I think emacs is fairly well suited for unit-tesing.

This is because, unlike many gui applications, things dont move around
very randomly in an emacs buffer. You can set-up a buffer to, for
instance, test font-locking. At a given character position you will
expect to find a certain property, if it works.

You have also the possibility to create macros, which re-create a
certain test environment, fairly conveniently with inbuilt emacs
facilities.

The problem is of course to get developers interested in creating
unit-test.

Another useful method is getting bug-reporters to supply unit tests
for their bugs.

If some unit-testing frameworks, like
http://www.emacswiki.org/cgi-bin/wiki/UnitTesting, where included in
emacs, maybe the bug-reporting template in emacs could suggest writing
a unit test for the bug when reporting it. 


> Of course, this is work and it takes time. But if the impression that
> changes create regression is valid then I think something like the
> above is a possible way to get out of it and get shorter release
> cycles.

-- 
Joakim Verona