Hi Max, I'm sure that is a leak related to sqlite WAL because if I disable WAL with journal_mode=DELETE and I do not have the leak. More over if I use vmmap tool I see the size of the mapped files growing up and vmmap show me that all the memory is used by -shm files mapped a lot of times. I'm using Windows XP 32bit, but even with win7 64bit I got the same result.
Il 09/09/2010 22.21, Max Vlasov ha scritto: > On Thu, Sep 9, 2010 at 7:16 PM, Dan Kennedy<danielk1...@gmail.com> wrote: > >>> Dan, don't know whether it is related, but I detected memory leak in >>> 3.7.2 >>> related to handle count increasing. >> I think there will be one open handle for each 32KB of >> shared-memory space in use. Or put another way, one open >> handle for every 4000 pages in the WAL file. They should >> all be closed when all connections to the database within >> the process are closed. >> >> >> > Dan, > thank you for your explanation, it makes perfect sense if we calculate the > sizes and everything. > > I tried to test till the sizes Michele mentioned, this time on Win7 64 bit, > I had almost perfect results, non-stopped appending till 1.7G and memory > size was never more than 50MB, so I'd suggest to Michele to check the code, > maybe there's a leak there not related to sqlite. > > But as a side effect I got not expected result in other area, when I tried > to append 1,500,000 records to this 1.7G file having 5G of free space on the > disk, I got the error a user reported recently about win7 64bit, "Disk I/O > error". (http://www.mail-archive.com/sqlite-users@sqlite.org/msg54935.html, > but this seems was not related to WAL) > > While I wrote this e-mail, I tried to perform another test, 1,200,000 > appends worked perfectly, the next series (without prior commit) failed > almost immediately with the same Disk I/O Error.Free size is still 3,7G, so > this was not related to the absence of free space. The size of shm file is > 0xAC0000, and the section starting 0xAB8000 till the end filled with zeros. > Please let me know if anything would be helpful to know from these files, I > keep it in the state they were after the failure, I even did not perform > Commit and Close. > > Max > _______________________________________________ > sqlite-users mailing list > sqlite-users@sqlite.org > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > > -- Selea s.r.l. Michele Pradella R&D SELEA s.r.l. Via Aldo Moro 69 Italy - 46019 Cicognara (MN) Tel +39 0375 889091 Fax +39 0375 889080 *michele.prade...@selea.com* <mailto:michele.prade...@selea.com> *http://www.selea.com* _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
