Reinoud Zandijk <zandijk@cs.utwente.nl> writes:
> Will this mean there will be new NetBSD 1.4.1 release-binaries with the
> good compiler and everything recompiled? I'm getting almost random
> kernel-crashes lately with the NetBSD/arm32 1.4.1 kernel and userland,
> especially when compiling under load. Has something to do with pv_map and
> pointers.

So, a bit of explanation about the NetBSD release engineering process:

Patch release binaries, e.g. the NetBSD/arm32 1.4.1 binaries, are
created from the sources of the patch release.  If the patch wasn't in
1.4.1, there's pretty much no such thing as "new 1.4.1 release
binaries" which will contain it.  There might be additional binaries
released with the patch, but, by virtue of the patch it wouldn't be
right to call them 1.4.1.

Problems like compiler bugs can easily be fixed in subsequent patch
releases: bug fixes get committed, and then "responsible parties,"
e.g. the port maintainer, the committer of the bug, or other
developers who feel themselves competent to do so can request that the
bug fixes be pulled up to the release branch.  They're then made
available as part of the release-branch bits up for FTP/SUP, are
accessable on the release branch via anon-cvs, and, finally, are
incorporated into subsequent patch releases.

What this means, really, is that if you've got a PR that's near and
dear to your heart -- e.g. compiler bug fixes -- which has been solved
in -current or could easily be solved in -current, you have to poke
the appropriate responsible parties (e.g. the port maintainer) to make
sure that the patches get applied and get pulled up to the release
branch.  Often, this happens without much prodding.  That's "The Right
Thing."  Other times, it takes more work.



cgd
-- 
Chris Demetriou - cgd@netbsd.org - http://www.netbsd.org/People/Pages/cgd.html
Disclaimer: Not speaking for NetBSD, just expressing my own opinion.