Обсуждение: Memory leak in large objects (was Re: Postgreqsl Large Objects)
Any ideas? -- Peter T Mount peter@retep.org.uk Main Homepage: http://www.retep.org.uk PostgreSQL JDBC Faq: http://www.retep.org.uk/postgresJava PDF Generator: http://www.retep.org.uk/pdf ---------- Forwarded message ---------- Date: Wed, 26 May 1999 08:41:26 -0400 (EDT) From: Brandon Palmer <bap@scl.cwru.edu> To: Peter T Mount <peter@retep.org.uk> Subject: Re: Postgreqsl Large Objects Ok, thanks, I will talk to him about the docs. Another problem that you may know about, want to know about. I have been using 6.5b for a while now with large objects. The problem that I am having is that I am running out of ram REAL FAST. I call a function that opens a given lo, then closes it (and some other stuff in the middle). When I do not open and close the objects, I am ok, but with them commented in, I have problems. I would think that it is not freeing all the memory from the open/close pair. I have talked to my LUG about this and they can not see any problems with the code that I have not. Take a look if you please: http://x.cwru.edu/~bap/search_3.c I get a few : NOTICE: LockReplace: xid table corrupted errors and then a NOTICE: ShmemAlloc: out of memory error. Thoughts? - Brandon (The only reason I am mailing this to you is that it seems there is a mem leak in the lo_open and lo_close functions)
Peter T Mount <peter@retep.org.uk> writes: > Ok, thanks, I will talk to him about the docs. Another problem that you > may know about, want to know about. I have been using 6.5b for a while now > with large objects. Beta1 you mean? There have been several LO bugs fixed since then, I believe. Please try it with a current snapshot. > I get a few : > NOTICE: LockReplace: xid table corrupted > errors and then a > NOTICE: ShmemAlloc: out of memory > error. Thoughts? I think this is memory corruption (something tromping on the shared memory allocation information), not a leak. regards, tom lane