I have uploaded a slightly different (and smaller) version of the script. It triggers the bug in both version 3.7.0.1 and 3.7.1 on my machine.
Thanks for testing, F. On Mon, Aug 23, 2010 at 3:00 PM, Black, Michael (IS) <michael.bla...@ngc.com> wrote: > I confirmed this shows a malformed result on 3.6.23.1 and 3.7.0.1 > > But it runs just fine under 3.7.1 > > So apparently whatever bug you triggered has been fixed now. > > > Michael D. Black > Senior Scientist > Advanced Analytics Directorate > Northrop Grumman Information Systems > > > ________________________________ > > From: sqlite-users-boun...@sqlite.org on behalf of Filip Navara > Sent: Mon 8/23/2010 7:43 AM > To: General Discussion of SQLite Database > Cc: d...@hwaci.com > Subject: EXTERNAL:Re: [sqlite] Help with database corruption? > > > > I can now reliably corrupt the database using standard commands. An > SQL script can be downloaded at the address below that creates a > database and then stresses it until a corruption happens. > > http://www.emclient.com/temp/sqlite_corrupt_log.zip > > Please help fix the problem or at least confirm that others can > reproduce it using the same script. > > Best regards, > Filip Navara > > On Fri, Aug 20, 2010 at 5:25 PM, Filip Navara <filip.nav...@gmail.com> wrote: >> Hello, >> >> is there anybody willing to help analyze corrupted database for >> possible bug in SQLite? >> >> It is a database file taken from one of our test machines and it is >> only few days old at most. The database file was only ever accessed >> with SQLite 3.7.0.1. It has page size of 1024, WAL mode, >> synchronous=FULL and incremental vacuum. No power failure happened on >> that machine. Apparently somewhere during the course of execution of >> our application the database free page list become corrupted. This is >> for the third time this week the same error happened, but only this >> time I have a copy of the database file. It is 887 Mb big (although >> the real data consume less than 2 Mb) and so it is rather problematic >> to provide the file. A copy is available at >> http://www.emclient.com/temp/mail_data.zip. >> >> The database was accessed on Windows machine using SQLite 3.7.0.1 >> 64-bit build. The following statements are the only ones that were >> executed against the database besides SELECTs and initial schema >> definition: >> >> INSERT INTO LocalMailContents ( >> "id", "partName", "contentType", "contentId", >> "contentDescription", "contentTransferEncoding", >> "contentMD5", "contentDisposition", "contentLanguage", >> "contentLocation", "partHeader", "partBody", >> "synchronizationKey", "contentLength") >> VALUES >> (@id, @partName, @contentType, @contentId, @contentDescription, >> �...@contenttransferencoding, @contentMD5, @contentDisposition, >> �...@contentlanguage, @contentLocation, @partHeader, @partBody, >> �...@synchronizationkey, @contentLength) >> UPDATE LocalMailContents SET synchronizationk...@synchronizationkey >> WHERE i...@id AND partna...@partname >> UPDATE LocalMailContents SET parthead...@partheader WHERE i...@id AND >> partna...@partname >> UPDATE LocalMailContents SET partbo...@partcontent WHERE i...@id AND >> partna...@partname >> DELETE FROM LocalMailContents WHERE id IN (<list of oids>); >> PRAGMA freelist_count; >> PRAGMA incremental_vacuum(<a number between 24 and 4096>); >> >> The error messages produced by "pragma integrity_check" are >> >> *** in database main *** >> Main freelist: invalid page number 866828 >> Page 396 is never used >> Page 473 is never used >> Page 14780 is never used >> Page 14915 is never used >> Page 153649 is never used >> Page 210894 is never used >> Page 319247 is never used >> Page 397006 is never used >> Page 416545 is never used >> Page 416636 is never used >> Page 416704 is never used >> Page 416705 is never used >> Page 416706 is never used >> Page 416707 is never used >> Page 416708 is never used >> Page 416710 is never used >> Page 416711 is never used >> ... >> >> I tried to locate the missing freelist pages in the database file and >> they definitely were there at some point, but I am not familiar enough >> with the file format to track the whole freelist and find where the >> corruption exactly happen. All I know is that page 388 is corrupted >> and points to a location outside of the database file. >> >> Thanks, >> Filip Navara >> > _______________________________________________ > sqlite-users mailing list > sqlite-users@sqlite.org > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > > > _______________________________________________ > sqlite-users mailing list > sqlite-users@sqlite.org > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
