* serv_getln now is a wrapper around existing functionality. a new temporary var in WC is used for performance reasons.
* non-buffered IO is going away completely, thus we remove the buffered suffix from the function and all places that used it so far
* set O_NONBLOCK on our server FD, hopefully this does what we want?
Art Cancro [Sun, 3 May 2009 03:38:07 +0000 (03:38 +0000)]
* Since we can now shut down the master SSH session cleanly with a command, there's no need to do the whole fork()/exec() thing. Replaced all that messy code with a simple call to system() and a SSH parameter to auto-background.
Art Cancro [Sun, 3 May 2009 03:30:35 +0000 (03:30 +0000)]
* Found a better way for ctdlmigrate to shut down the master SSH socket. SSH has a command to do it. No more silly games trying to hunt for the correct pid to kill.
* temporary solution to our ser_readln migration: have a buffer on the session, read there to, and copy it over then.
* remove some more serv_readln's, add some more GetServerStatus'es
* start migrating important message to strbuf; have both meanwhile.
* evaluate result of gotoroom; this needs to be done elsewhere too.
* fix roomname thing in useredit
Art Cancro [Sun, 26 Apr 2009 03:46:53 +0000 (03:46 +0000)]
* Use fork() and exec() instead of system() to set up the master SSH session, so we can get the pid to kill later
* At exit, kill the master SSH session and delete the socket
* remove old notes rendering code, its not used anymore.
* add mail attachment inline renderer for notes
* add first draft template stripped down from the original notes thing, needs some more love.
Art Cancro [Fri, 24 Apr 2009 15:14:43 +0000 (15:14 +0000)]
* sendcommand default watchdog timeout set to 60 seconds. 5 seconds is way too low.
* ctdlmigrate overrides the default and sets it to 3600 seconds on both the source and target hosts.
Art Cancro [Fri, 24 Apr 2009 02:09:30 +0000 (02:09 +0000)]
* Added a subcommand to the migrator to list out all of the directories which require copying because they are not part of the database; these can be fed into rsync to supplement the XML dump and build a complete replica of a Citadel installation.
Art Cancro [Wed, 22 Apr 2009 02:25:02 +0000 (02:25 +0000)]
* Handle wandering punctuation in charset definitions. This fixes the most common WebCit crash being experienced since the 7.50 release. Committing to both trunk and stable-75x.
* persistantly handle the bufferlengths for dates; snprintf doesn't terminate strings if they exceed the provided length, which was causing character soup in some rare cases.
* give all commands their own function
* evade rumpelstilskin-lookup in favour of module registration commands
* replace linked list of commands by 4-char-hash.
Art Cancro [Thu, 16 Apr 2009 21:24:32 +0000 (21:24 +0000)]
* Made our Berkeley DB errors look more like the ones thrown by other applications, so that when people Google them they don't get led to a 'Citadel Sucks' thread on a popular Ubuntu support forum that was populated by paid Zimbra astroturfers.
Art Cancro [Thu, 16 Apr 2009 14:29:18 +0000 (14:29 +0000)]
* Completed optimization of load_ical_part(), now uses the data supplied by MSG4 if it is usable. In initial testing this cuts calendar load time almost in half.
Art Cancro [Tue, 14 Apr 2009 14:31:00 +0000 (14:31 +0000)]
* Adjust the memory allocation for Base64-converted messages in serv_migrate.c ... I guess I shouldn't have overconfidently used the words 'well-tested formula' in the comment next to it.
* add fast linebuffered function just cutting its buffer before it reads a new chunk; this way we save lots of memmoves when reading large line arrays line by line.