Art Cancro [Mon, 22 Feb 2010 18:30:51 +0000 (18:30 +0000)]
* Renamed the 'jabber' module to 'xmpp' module. Also renamed functions starting with jabber_ to xmpp_ . Obviously I did *not* rename any of the XML namespaces because that would break the protocol.
* since valgrinds name reacting to killall has become verry fancy to type, do the same on SIGINT as SIGHUP when running from the console: cleanup shit and exit.
* don't break overlong lines on random places and loose chars, find the nearest previous / next whitespace and kick it there.
* don't use cprintf, since we know the buffersize hand have reserved space for the newline... client_write is faster.
Art Cancro [Wed, 17 Feb 2010 17:29:05 +0000 (17:29 +0000)]
* text client: try the user's password with and without strproc() processing. This will allow authentication of users in LDAP etc. who have special characters in their passwords, without breaking the sites using self-contained auth that have passwords stored (unbeknownst to the users) whose passwords went into the db stripped. Resolves bug 457.
Art Cancro [Wed, 17 Feb 2010 16:53:03 +0000 (16:53 +0000)]
* Switch from ldap_search_st() to ldap_search_ext_s(). The latter is not deprecated.
* Ignore the return value from ldap_search_ext_s() because it appears to be misleading. Instead, check to see whether the search returned any results (null or not null). This also fixes a potential memory leak resulting from ldap_search_ext_s() returning an error code but still populating the search results.
Art Cancro [Mon, 15 Feb 2010 22:47:57 +0000 (22:47 +0000)]
* Implemented a new way to show where the new and old messages are, that is visually consistent with the rest of the layout. TRANSLATORS: there are a few new strings; please localize
* don't flush the inbuffer on send; while this is appropriate for a citadel client, it may irretate smtp traffic
* add some debug code for the linebuffered reader so we may find it again later
* fix the linebuffered reader; some cases (like having a part of a line in the buffer, and reading more of the same line) didn't work
* if we have the chance to flush the IOBuf because of we read everything inside, take it!
* if we delete the last message, msglist[num_msgs - 1] points into an invalid area
* qrbuf.QRhighest is set to 0 if the room is empty; this will most probably fix the problems with empty rooms appearing to have unread messages.
Art Cancro [Sat, 6 Feb 2010 20:08:35 +0000 (20:08 +0000)]
* Ok, this is it: the 'new new' bbs view is now feature complete. Messages are displayed in a simple paginated format. When the selected operation is 'read new' it goes to the page containing the first new message and then automatically scrolls down to that message.
* move serv_func.c:read_server_binary() to tcp_sockets.c: serv_read_binary()
* serv_read_binary() now knows a way to get the whole thing at once instead of scattering in 4k blocks
* close the server file in all conditions while loading images, else our facility might block in the server
* add -R commandline switch so the user can limit the blobsize if the tcp connection citserver <-> webcit is expected to be less stable
* remember the total size a downloadable file has
* use mmap to read the download file for output; this way we don't need to copy it into memory first and can let the kernel do this job
* remove the 10k transfer limit
Art Cancro [Fri, 5 Feb 2010 05:38:39 +0000 (05:38 +0000)]
* Greatly simplified the bbsview code by eliminating the concept of using a 'pivot point' where the new messages start. What I'm checking in now is just a simple paginated view where the page numbers start at zero and all contain the same number of messages. Tomorrow I will make 'read new' work correctly by making it jump to an anchor within one of those pages.
* remove unneeded vcard function...
* put the mime struct through into the vcard functions instead of just parts
* evaluate the (new) charset from part= and DLAT
Dave West [Sat, 30 Jan 2010 16:11:50 +0000 (16:11 +0000)]
Attempt to fix the crash that results from a new sessions context
not existing before the command loop is entered.
This fixes a theoretical non atomic opperation but does not seem
to cause unwanted side effects. Lets see if it cures the problem
on those systems that are seeing it regularly.
* merge from dav_rework:
* modularize the parser for load_msg_ptrs
* add parser that has the euid
* add work on the URL parsers
* unfinished: new propfind; is #ifdefed
* merge in much of the dav_rework branch: (the drift was becoming to big, no functional changes)
* make configure search for cunit
* rename struct floor to Floor, since floor was already taken by math.h
* move over most of the functions from webserver.c to sysdep.c so main() is isolated and more easy to be replaced by another main in a test-suite-binary
* add first unit test implementations
- test the LANG header parser
- start testing the floor evaluation