On Tue, 20 Dec 2016 at 09:20:05 -0500, Jeremy Bicha wrote: > On Tue, Dec 20, 2016 at 12:55 AM, Michael Biebl <[email protected]> wrote: > > It's most certainly related to the new seccomp based sandbox. > > Michael, the three latest commits on tracker master from December 17 > (not commited to the 1.10 branch) look significant too.
The specific crash reported in #848842 is in mlock() as used by libfluidsynth, for which see https://bugzilla.gnome.org/show_bug.cgi?id=776117 and upstream commit c9acfe0e3a3ee5b809860845b856a876c7d42eb0 "Handle mlock*/munlock* syscalls" (which was added after those to which you refer). 73f5cd54a2e8b4ea8fbb4b3997b24d5e4aebd671 "Whitelist umask" doesn't have a Debian bug report that I can immediately see, but seems harmless and a good idea to have. 6a05068624bfa55233e2c7e107f5749a7ad4deb8 "Disable libmediaart dependency" is entirely reasonable upstream, but unnecessary if we --disable-libmediaart, AFAICS. 435eca3325d2e27183de2e12c3d4cc840e362fb1 "Whitelist dup/dup2/dup3" is https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=849936 in Debian and https://launchpad.net/bugs/1649035 in Ubuntu. > We might want to use --disable-libmediaart If tracker-extract is going to use the seccomp sandbox in stretch (which I personally think it should), then it can't use libmediaart, because libmediaart does write to disk. > On Ubuntu there were also crashes from getrusage() > https://launchpad.net/bugs/1649004 That's 47b4958fcdbf758e4a90ecc06e58e321372bef18 "Allow querying process stats/limits", I think. Building a test version that has all of those. S
