I tried the experiment again with -g3 -O0, I got less information than expected (there are still many unknown symbols in libsqlite3.so), but the function requiring all this memory is sqlite3_step.
So maybe it is one complex query? I would like to avoid excessive swapping on the production server, maybe I should not worry at all? On Tue, Jul 16, 2013 at 1:11 PM, Paolo Bolzoni <paolo.bolzoni.br...@gmail.com> wrote: > On Tue, Jul 16, 2013 at 1:00 PM, Dan Kennedy <danielk1...@gmail.com> wrote: >> On 07/16/2013 01:49 AM, Paolo Bolzoni wrote: > >> A very large blob or string result? > I would exclude this, I do use blobs... but they are at most > few dozen of bytes... > >> Code allocates (or leaks) >> tremendous numbers of sqlite3_stmt* handles? > Thanks to RAII the code should not leak (also valgrind > confirm this). I allocate statements, but I deallocate only > at the end. So it cannot explain a peak in memory usage. > >> SQLite has various APIs for querying memory usage: >> >> http://www.sqlite.org/c3ref/memory_highwater.html >> http://www.sqlite.org/c3ref/c_status_malloc_count.html >> >> Or, using the shell tool, the ".stats" command can be used >> to access the same values. > I guess I can see something. > > At the moment I am running the test again using a sqlite3 > version compiled with -g3 and -O0 so I hope I can get more > insight... _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users