Issue 95: Correctly handle inotify_add_watch error

Currently recollindex simplely quit is one of the inotify_add_watch failed. Even if the error is just a Permission denied. I think it’s silly to stop whole index process for a single failure, so please handle the error correctly.

I’ll admit that unconditionnally exiting is a bit extreme here… If you can test the attached patch, please tell me how it works for you. It should be in the next release anyway.

jf

Thanks, it works

The patch doesn’t resolve the problem for me, I have tried using also the latest version: {{{ Recoll 1.18.0 + Xapian 1.2.3 }}}

The command I used to launch recollindex was: {{{ time /usr/local/bin/recollindex -m -w 1 -x -D }}} for 90Gbytes of data.

And in some computers the daemon still dies after first indexing…using debug=4 I got: {{{ … :4:../rcldb/rcldb.cpp:1498:Db::needUpdate:no: [Q/ABSOLUTE/PATH/TO/DOCUMENT5.doc|] :4:../rcldb/rcldb.cpp:1498:Db::needUpdate:no: [Q/ABSOLUTE/PATH/TO/DOCUMENT6.doc|] :4:../rcldb/rcldb.cpp:1498:Db::needUpdate:no: [Q/ABSOLUTE/PATH/TO/DOCUMENT7.doc|] :4:../rcldb/rcldb.cpp:1498:Db::needUpdate:no: [Q/ABSOLUTE/PATH/TO/DOCUMENT8.doc|] :4:../rcldb/rcldb.cpp:764:Db::i_close(0): m_isopen 1 m_iswritable 1 :4:../rcldb/rcldb.cpp:774:Rcl::Db:close: xapian will close. May take some time :4:../rcldb/rcldb.cpp:779:Rcl::Db:close() xapian close done. :4:../internfile/mimehandler.cpp:97:clearMimeHandlerCache() :4:../utils/execmd.cpp:141:ExecCmd: killpg(1971, SIGTERM) :4:../utils/execmd.cpp:141:ExecCmd: killpg(1735, SIGTERM) :4:../utils/execmd.cpp:141:ExecCmd: killpg(3316, SIGTERM) :4:../utils/execmd.cpp:141:ExecCmd: killpg(1628, SIGTERM) :4:../utils/execmd.cpp:141:ExecCmd: killpg(1736, SIGTERM) :3:rclmonprc.cpp:309:RclMonEventQueue: not ok: queue terminated :4:rclmonprc.cpp:571:Rclmonprc: calling queue setTerminate :4:rclmonprc.cpp:575:Monitor: returning :4:../rcldb/rcldb.cpp:649:Db::~Db: isopen 0 m_iswritable 0 :4:../rcldb/rcldb.cpp:764:Db::i_close(1): m_isopen 0 m_iswritable 0

real 45m33.938s user 20m17.028s sys 12m55.788s }}}

I replaced on this output the original path and documents names with: {{{ /ABSOLUTE/PATH/TO/DOCUMENTx.doc }}}

recoll.conf contains: {{{ indexstemminglanguages = english spanish topdirs = /ABSOLUTE/PATH/ monixinterval=600 filtermaxseconds=300 compressedfilemaxkbs=20000 loglevel=4 daemloglevel=4 }}}

Difficult to be absolutely sure that this is the same problem from seeing the log, but we’ll solve it anyway, if you can work with me.

I can see a few cases in the code where the thread which receives the events could call for a stop without printing a message. Here is what I suggest: - Please build from the 1.17 branch, the tip (1.18) is undergoing changes right now, and the index format is incompatible, so lets' stay with the "stable" code. Use either the 1.17.3 tarball or the 1.17 branch from hg. - I am attaching 2 modified source files (which go in the index/ directory). The modifications are only additional messages in hope that we’ll get an explanation about why the indexer is exiting.

Sorry, wrong versions for the modified files. Attaching new ones real soon now…

Fixed in current trunk code, closing.