Upgrading zlib to the latest release (1.2.5) fixed all of my valgrind warnings in Fossil, including those that appeared to have been coming from SQLite WAL.
On Fri, May 13, 2011 at 3:11 PM, Richard Hipp <d...@sqlite.org> wrote: > > > On Fri, May 13, 2011 at 2:49 PM, Richard Hipp <d...@sqlite.org> wrote: > >> >> >> On Fri, May 13, 2011 at 2:42 PM, Steven Parkes <smpar...@smparkes.net>wrote: >> >>> I'm getting conditional branch warnings from valgrind that all percolate >>> up via a WAL path. Anybody know if these are known & benign? >>> >> >> I noticed this myself just yesterday. >> >> On our pre-release checklist, we have a bullet to run a full branch >> coverage test under valgrind with no unexplained warnings. (Some of the >> test cases in our test suite deliberately do things that upset valgrind - >> but those cases are all well-marked.) And we did this on the most recent >> releases with no problems. But, I've been running that test on 32-bit >> linux. And the valgrind warnings are only showing up on 64-bit linux. At >> least, they only show up on 64-bit linux for me. I reran the test on 32-bit >> linux just to double-check, and sure enough, there were no warnings. >> > > Correction: It wasn't the SQLite 100% branch cover test where I saw this > problem yesterday; it was a test run of Fossil, specifically: > > valgrind ./fossil rebuild > > We also have a bullet to run fossil using valgrind as part of the > pre-release testing. And as I was doing a fossil release yesterday, that > was the test I was running. The errors are being reported out of the WAL > code in SQLite, which is embedded in Fossil. And the errors were only > showing up on linux-64. The 100% branch coverage tests for SQLite are > coming up clean in valgrind on both 32-bit and 64-bit linux. > > Valgrind also gives a bunch of warnings on the zlib compression library on > 64-bit linux that it is quiet about on 32-bit linux.... > > >> >> I have not yet figured out why valgrind warns on linux-64 but not on >> linux-32 and whether or not this is anything to be concerned about. >> >> >>> >>> This is, admittedly, off a sqlcipher-patched version of 3.7.6.2. I don't >>> think that should relate but I'll check if they shouldn't be occurring. >>> _______________________________________________ >>> sqlite-users mailing list >>> sqlite-users@sqlite.org >>> http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users >>> >> >> >> >> -- >> D. Richard Hipp >> d...@sqlite.org >> > > > > -- > D. Richard Hipp > d...@sqlite.org > -- D. Richard Hipp d...@sqlite.org _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
