diff options
author | http://joeyh.name/ <http://joeyh.name/@web> | 2013-08-26 19:51:48 +0000 |
---|---|---|
committer | admin <admin@branchable.com> | 2013-08-26 19:51:48 +0000 |
commit | 51152afee831b76f9c79a2d9e4be30b61cb8cd7e (patch) | |
tree | 6b5fd0b2861317d50649d0e4d6ba71cde79ef975 /doc | |
parent | 4fb2080abe5d94e6dcbf88d1862d0643d0424332 (diff) |
Added a comment
Diffstat (limited to 'doc')
-rw-r--r-- | doc/bugs/git-annex_quit_unexpectedly___40__macosx__41__/comment_1_97abb8442329d19c9687002f43afac74._comment | 23 |
1 files changed, 23 insertions, 0 deletions
diff --git a/doc/bugs/git-annex_quit_unexpectedly___40__macosx__41__/comment_1_97abb8442329d19c9687002f43afac74._comment b/doc/bugs/git-annex_quit_unexpectedly___40__macosx__41__/comment_1_97abb8442329d19c9687002f43afac74._comment new file mode 100644 index 000000000..4dcc5412c --- /dev/null +++ b/doc/bugs/git-annex_quit_unexpectedly___40__macosx__41__/comment_1_97abb8442329d19c9687002f43afac74._comment @@ -0,0 +1,23 @@ +[[!comment format=mdwn + username="http://joeyh.name/" + ip="4.154.0.63" + subject="comment 1" + date="2013-08-26T19:51:48Z" + content=""" +So this is a crash in the gnutls library used for XMPP. Someone else using OSX reported a similar crash to me by email, reproducible reliably by setting up xmpp with google. This is great debugging info: + +<pre> +Thread 3 Crashed: +0 libsystem_c.dylib 0x00007fff86ae0bf9 pthread_mutex_lock + 20 +1 H 0x000000010e9fd29f gnutls_system_mutex_lock + 12 +2 H 0x000000010ea7fa29 wrap_nettle_rnd_refresh + 20 +3 H 0x000000010e9fee89 gnutls_deinit + 42 +4 git-annex 0x000000010caf0a3a 0x10b5e6000 + 22063674 +</pre> + +Looks like `gnutls_deinit` was called and it crashed there, inside pthread code. So I think git-annex has already managed to connect to the XMPP server (to test it) and the cleanup is where it's crashing. + +This is the second time I have seen a gnutls-related crash using XMPP. The other one was tracked down by John Millikin to a resource allocation bug in haskell-gnutls and fixed. + +I have written a test case that reproduces the problem -- just forking a dozen threads that each try to connect to the google xmpp server and then close the connection. After a dozen or so succeed, one will reliably cause a segfault. Forwarded this test case to John. +"""]] |