Jamie, Here comes the patch for this crash, fixed on Solaris, can search numbers now. Halton. On Sun, 2007-09-23 at 22:35 +0800, Halton Huo wrote:
Use '123' in tracker-search-tool will core dump, the debug information is as following. Finished optimizing. Waiting for new events... ** CRITICAL **: file tracker-indexer.c: line 1432: assertion `query->words' failed aborting... t 2 (l 2) signal ABRT (Abort) in __systemcall6 at 0xd0e40d66 0xd0e40d66: __systemcall6+0x0006: jae __systemcall6+0x18 [ 0xd0e40d78, .+0x12 ] Current function is tracker_get_hit_counts 1432 g_return_val_if_fail (query->words, NULL); (dbx 17) p query->words query->words = (nil) (dbx 18) where current thread: t 2 [1] __systemcall6(0x3, 0xffbffedf, 0xfff7, 0xd0e87000, 0xd0a099b4), at 0xd0e40d66 [2] __lwp_sigmask(0x3, 0xd08f0360, 0x0), at 0xd0e1e2ff [3] _thr_sigsetmask(0x2, 0xd0a099c0, 0x0), at 0xd0e334b8 [4] _sigprocmask(0x2, 0xd0a099c0, 0x0), at 0xd0e33549 [5] _sigrelse(0x6), at 0xd0e1f50a [6] abort(0x8104770, 0x84b2218, 0xd0ff7424, 0xd0de95cd, 0xd10a4da1, 0x0), at 0xd0dd4fb6 [7] g_logv(0x0, 0x8, 0x80db7ac, 0xd0a09edc), at 0xd0f6f9ec [8] g_log(0x0, 0x8, 0x80db7ac, 0x8100cf0, 0x598, 0x8100d04), at 0xd0f6fa15 =>[9] tracker_get_hit_counts(query = 0x858a200), line 1432 in "tracker-indexer.c" [10] tracker_dbus_method_search_get_hit_count_all(rec = 0x8547038), line 169 in "tracker-dbus-search.c" [11] process_user_request_queue_thread(), line 1662 in "trackerd.c" [12] g_thread_create_proxy(0x84b2218), at 0xd0f88caa [13] _thr_setup(0xd08f0200), at 0xd0e3d962 [14] _lwp_start(0x3, 0xffbffedf, 0xfff7, 0xd0e87000, 0xd0a099b4, 0xd0e331fe), at 0xd0e3dbc0 On Sun, 2007-09-23 at 14:12 +0100, jamie wrote:Hi All, the next version of tracker is sitting in svn (there are still a few bug fixes to apply but should be stable for the most part) this version implements index merging which dramatically improves speed and scalability of indexing (as well as eliminating fragmentation and disk IO issues0. 250 MB of linux kernel source was indexed in less than 8 minutes so it should index large amounts in very short time) because of that there are some major changes in the way things ar e handled so it really needs a lot of thorough testing before release we use by default a 16MB word cache as each flush produces a separate mini-index which are later all merged together to form the main index. Search results will only be available after the first flush and no new search results will be added from subsequent flushes until merging is complete so please can you test the following: 1) index as normal (no need to use --reindex - it should reindex itself auotmatically) 2) keep an eye on memory usage - there may be a few leaks but it should plateau around 30mb 3) after index test makking changes to files and moving stuff around 4) do source checkouts so that lots of files are overwritten in one go and pay attention to speed/slowdown of system 5) create lots of new files and make sure its fast and new results show up pls also run with -f in gdb ss thta you can get me backtraces for non-crash errors (as well as crashed ones!) if all goes well I will release tomorrow... thanks jamie _______________________________________________ tracker-list mailing list tracker-list gnome org http://mail.gnome.org/mailman/listinfo/tracker-list_______________________________________________ tracker-list mailing list tracker-list gnome org http://mail.gnome.org/mailman/listinfo/tracker-list
Attachment:
tracker-05-crash-on-number.diff
Description: Text Data