RE: too much untested code in new kernels

Jon Lewis (linuxsys@ssg.com)
Tue, 7 Jan 1997 20:18:26 GMT

William Burrow writes:
> > It's clear that a volunteer effort with minimal funding can't be
> > expected to deliver the sort of rigorous testing one hopes a major
> > company would do but there are all too many posts to this list which,
> > in effect, say "oh, silly me, make the ':' on line 102 a ';'". These
> > are not subtle errors or problems of interaction but problems with
> > typographical errors and misspelled variables, etc.
>
> Perhaps if the developers themselves do not have the time to go through
> their own code or test it, a team of Linux dev testers could be formed?
> The idea would be to stress, in particular, those sections of the kernel
> that had been changed, rather than to attempt to use the kernel for
> general use.

In a sense, anyone who runs the 2.1.xx tree is in the class of "Linux
dev testers" but I agree that there is a need for people who have the
resources to test the more complex hardware combinations for otherwise
unapparent problems.

> > Before someone thinks I'm dumping this all in Linus' lap, I'm not.
> > It's clear he's got his hands full and the rest of the developers need
> > to take up the load he can't be expected to handle with the present
> > demands on his time.
>
> Some seem to take a lackadaisical attitude towards these untested code
> type bugs, but this type of bug can catch up to you later. Code that is
> known to work at various stages may still break due to subtle
> interactions later, but at least it isn't subtle interactions with broken
> code.

Bingo! It makes the task of debugging legitimate problems (e.g.,
unexpected races) doubly difficult if the job also involves cleaning
up typos and obvious errors.

Rick