Ah, mentioning PerlIO and DBD::Oracle points the finger at the fact that DBD::Oracle hasn't been updated to use PerlIO for warnings/trace (at least not in the release on CPAN). That may be a factor.
If DBD::Oracle needs to 'say' anything (which is very rare in normal operation) it'll be passing a PerlIO pointer to stdio functions. Any kind of damage could result. Tim. p.s. I'm about to release DBI 1.33 and I now have a new Sun box I can setup Oracle on so there's hope I can get back to DBD::Oracle development after an uncomfortable long period of, er, stability :) On Wed, Feb 26, 2003 at 11:10:43AM -0600, Neibarger Scott H wrote: > I'm afraid I can't offer any feedback that would be useful, other than the > fact that Lincoln is most likely correct that this problem is manifested due > to changes in PerlIO. I have several daemons running on Tivoli TMR servers, > many of which use DBI and DBD::Oracle; one of which is experiencing some > serious issues with memory leaks and internal corruption. I've scanned the > perlpod documentation for 5.8.0 only to find that there're several entries > mentioning memory leaks (in lexicals and anonymous subs) and internal > corruption (unsafe handling of signals). I'm in the process of preparing to > upgrade Perl from 5.005_03 to 5.8.0, which I'm hoping will solve the core > dumps and internal corruption I'm seeing in one of the daemons that performs > significantly more I/O operations than the others. > > I would like to take this moment to thank Lincoln Stein for his wonderful > book "Network Programming with Perl"; it's been a wonderful resource for me > on this current project. If you've not purchased it, support Dr. Stein and > buy his book! > > Scott H. Neibarger > SRA International, Inc. > IRS ESM project > 512-460-8717
