On Sat, 28 Apr 2001 23:53:05 -0400,
"Eric S. Raymond" <[EMAIL PROTECTED]> wrote:
>Michael Elizabeth Chastain <[EMAIL PROTECTED]>:
>> This would be a very useful feature for CML2 as well. With a random
>> configuration generator, anyone can step up and build 100's of
>> configurations, which catches a lot of errors.
>
>CML2 would probably reject most random configurations for violating some
>consistency constraint or other. What kind of errors eould we be hunting,
>anyway?
Not raw random configurations, the idea is to generate a random config
then run it through make oldconfig (CML1) or any CML2 to generate a
consistent config. Then try to compile and link using that config, the
aim is to catch broken makefiles or headers, like the bug reported in
http://www.mail-archive.com/linux-kernel%40vger.kernel.org/msg41684.html
In my case I want to stress test the 2.5 Makefiles to see if I covered
most of the cases. Hand converting from the 2.4 mess is bound to have
introduced some problems.
_______________________________________________
kbuild-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/kbuild-devel
