Tom Lane wrote:
This is not the same test data as in your previous concurrent-index
problem, then? I still had a copy of that, so I tried the case, and
it doesn't crash on that data ...
Teodor Sigaev wrote:
I tryed to reproduce the bug but without success.
Could you provide a dump of text column?
It's not the same data, no. It's about 7 times as large as the old test
data. I'd upload it as well, but I'm limited to 8KB/sec upstream at the
moment, which makes it a bit of pain.
However, I may not be duplicating your dictionary exactly. I didn't do
anything with the compound.pl script in that tarfile --- was I supposed
to?
No, compound.pl is already applied to the files in the archive. I've
only converted them to UTF-8. I've used this dict with 8.2.x for quite a
while in production.
Maybe it's faster when I setup a virtual machine and grant SSH access to
it. The problem is that I'm running out of time. I'm probably not able
to do it before friday :(
Is there anything quick I can do? Can I make the backtrace more
meaningful by dropping -O2 or add -ggdb? Add some debug to SplitToVariants?
--
Best regards,
Hannes Dorbath
---------------------------(end of broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster
