On Tue, Jun 19, 2001 at 12:37:20PM +0100, Dave Hodgkinson wrote:
> Matthew Byng-Maddick <[EMAIL PROTECTED]> writes:
> > On Tue, Jun 19, 2001 at 12:01:59PM +0100, Dave Hodgkinson wrote:
> > > Matthew Byng-Maddick <[EMAIL PROTECTED]> writes:
> > > > I'm trying to keep the thing stable (I knew I couldn't run away from
> > > > sysadmin for long...), and the board hits MySQL to the extent that all
> > > > those bugs actually get to manifest themselves...
> > > Are you on MySQL 3.23? On BDB?
> > > If you're doing high-update frequency stuff on anything less you get
> > > what you deserve and have no right to whinge.
> > 3.23 Yes, BDB no.
> > Actually some of the bugs were crashes in the threading under FreeBSD 4.
> > :-(. The bug would manifest itself as a crash in the return() call of
> > mysql_select() :-(
> > MBM (may well try BDB and see if it makes any difference...)
> If it's threading that's the issue, then unlikely.
well, I think it manifests itself when you've got writes and reads, and
some semaphore doesn't get set properly. I'm not exactly sure, because
the code is incomprehensible, and you shouldn't *ever* get a SEGV in
return.
> FreeBSD? Ooooh. Is there a tarball release for that? Do you build it
> yourself?
The ports collection has a MySQL 3.23 port, but I have also tried building
it myself to see if it helped. It didn't.
Now, we've got one running under linux-emulation in FreeBSD, as we couldn't
persuade it to compile without threads, and for some extremely crap reason,
the parent's wait() call gets ignored, so we seem to end up with large
numbers of zombies. :-(
I'm at the end of my tether with this...
> Certainly under linux, they statically link to a patched libpthreads.
Right, I'm half inclined to try *that* now....
MBM
--
Matthew Byng-Maddick <[EMAIL PROTECTED]> http://colondot.net/