Hi all. I wrote a summary way back in November about how I got mysql to
compile with this configure string on Tru64 4.0D:
-----------
env CC="cc -pthread" CFLAGS=-I./include CXX=gcc CXXFLAGS=-O3 ./configure \
--with-low-memory --with-named-thread-libs="-lpthread -lmach -lexc"
----------
It compiled fine, but we've had trouble with it crashing and dropping
connections with even a moderate load. I also discovered that the test
scripts in the distribution directory didn't do well. I tried tests with
fork_test.pl and fork2_test.pl, as suggested in the manual if crashes
happen. They tended to fail after only a few iterations. Luckily the
startup script safe_mysql would immediately restart it when that happened,
but it was still a big problem.
I have now upgraded to 3.22.30 from 3.22.27, but more importantly I
believe, I have upgraded gcc to 2.95.2 from 2.8.1, following advice in the
archives of the mysql mailing list which I hadn't found before. There is a
bug in the compiler that shows up when you do that, but the manual tells
you how to get around it in the osf compile problems section. Our mysql
server is much more stable now, and the fork test scripts can do their
originally configured 10000 interations with no problem. I highly
recommend a gcc upgrade and maybe server upgrade too if you are having
problems and you have used gcc for a part of the compile as above.
Also of note, there is a bug that was brought up in BugTraq for
pre-2.29.30 (or possibly pre- 2.29.29 or 2.29.28) versions, that makes all
mysql passwords available to a careful exploit run by a user who starts
out knowing just one password. If you search recent BugTraq for mysql,
you can find a patch there if you don't need to upgrade otherwise.
Best,
Ann Cantelow
cantelow_at_athena.csdco.com
Received on Wed Jan 26 2000 - 18:58:17 NZDT