Sorry been away a while.
The context re-work branch has most of this already done. You're basically
repeating what I've done.
MasterCC still exists but only until we make the modules use the newer
method. Ofcourse we need to merge that branch in first.
Tue Jun 08 2010 17:30:41 EDT
Sorry guys, can't be much help to you at the moment.
I'm totally burried in work stuff.
Should get better in next week or so though.
dothebart, check the thread structure for something containing the thread id.
the macro CT will always return a pointer to the structure for the current
Ah, thought we'd done the 7.80 release already.
Sat Apr 17 2010 15:24:48 EDT from IGnatius T Foobar @ Uncensored
davew: this merge is going to have to be post-7.80 ... and you're going to
be the one doing the merging, so please keep your branch up to date. We
were supposed
Hmm, That creates a warning on my system.
My version of db.h expects int (*)(void)
I'll try upgrading.
Sat Mar 27 2010 07:07:37 EDT from dothebart
dotheb...@uncensored.citadel.org Subject: Citadel commit log: revision
8496
r8510 | davew | 2010-04-11 16:20:40 -0400 (Sun, 11 Apr 2010) | 3 lines
Changed paths:
M /branches/citadel_context_rework/context.h
M /branches/citadel_context_rework/database.c
M /branches/citadel_context_rework
OK guys I reckon my branch is ready to merge with the trunk.
It changes a lot of things so you will realy need to get familiar with it
Changes are generally documented where they occur if its not obvious from the
function name or useage.
My branch contains a modified pop3client and a
I came to the same conclusion.
In my branch there is a big comment at the top of the file stating how
horrible it is to the system.
Sat Mar 27 2010 07:31:21 AM EDT from dothebart @ Uncensored Subject:
cmd_chat
reading through that function, its blocking one context and FD,
Sorry guys. Missed the weekend of test and fix. I had to go to Scotland and
fix an Ice rink.
OK,
So who's about on IRC this weekend. We'll get some testing and bug fixing
done.
Can you change that to SIG_USR2. I'm using USR1 to interrupt select when the
list of new contexts changes.
Sun Mar 14 2010 06:31:08 PM EDT from dothebart
dotheb...@uncensored.citadel.org Subject: Citadel commit log: revision
8456
Hmm, SMTP outgoing fails? since when? I'm running my branch server and it
seems to be working fine and its merged with the trunk HEAD (more or less).
Sun Mar 14 2010 11:44:37 PM EDT from IGnatius T Foobar @ Uncensored
Subject: uncensored reverted to old code
Ok, due to too many
Hmmm, I cant find a single strchr in the previous revision of the code you
changed. I can see a strstr that you have eliminated though. I take it that
is what you mean.
Anyway,
looks good to me. I've not fed anything to it yet though.
Tue Mar 02 2010 18:07:54 EST from
OK, given that can we change the description in webcit to be something more
meaningful for what it does.
that is change this
Server connection idle timeout (in seconds)
to something like
Dead client connection time out (in seconds)
The current infers that it is there to time out
r8398 | davew | 2010-02-28 09:24:54 -0500 (Sun, 28 Feb 2010) | 3 lines
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/context.h
Seems to be stable.
r8401 | davew | 2010-02-28 10:31:40 -0500 (Sun, 28 Feb 2010) | 12 lines
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/modules/calendar/serv_calendar.c
M /branches
r8402 | davew | 2010-02-28 11:04:50 -0500 (Sun, 28 Feb 2010) | 2 lines
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/modules/chat/serv_chat.c
M /branches
r8403 | davew | 2010-02-28 11:36:09 -0500 (Sun, 28 Feb 2010) | 3 lines
Changed paths:
M /branches/citadel_context_rework/context.c
Clean a few debug messages.
r8410 | davew | 2010-02-28 15:30:00 -0500 (Sun, 28 Feb 2010) | 5 lines
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/include/ctdl_module.h
M /branches
I've just about completed the context rework I've been working on and am
almost ready to merge it into the HEAD.
I would like you all to run as many tests on it as you can.
If you've not been following the commit logs or just can't remember them all
there are lots of changes.
Several of
as defines..
and just found out we were using numbers all over the place... done.
Dave, whats wrong with memreadline? me making it const? should we have two
of them?
Fr Feb 26 2010 14:05:27 EST von davew @ Uncensored Betreff: Citadel
commit log: revision 8384
Oh great another
Ah, you spotted it.
Sat Feb 27 2010 11:01:45 AM EST from davew @ Uncensored Subject:
Citadel commit log: revision 8384
prototype in libcitadel.h is const but defenition in tools.c isn't. Stops my
build.
Sat Feb 27 2010 05:33:31 AM EST from dothebart @ Uncensored Subject
If I've got this right. It means the preprocessor checks one has used a valid
value thus preventing spelling mistakes?
Sat Feb 27 2010 07:46:23 AM EST from dothebart
dotheb...@uncensored.citadel.org Subject: Citadel commit log: revision
8392
r8396 | davew | 2010-02-27 18:48:18 -0500 (Sat, 27 Feb 2010) | 5 lines
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/modules/network/serv_network.c
M /branches
Question time.
the context contains lastcmd and lastidle.
lastcmd contains the time the command loop last got run.
lastidle contains the time of the last command that was not one of NOOP,
QNOP, PEXP, GEXP, RWHO, TIME
client connections are timed out based on lastcmd which means they
r8380 | davew | 2010-02-26 06:01:38 -0500 (Fri, 26 Feb 2010) | 3 lines
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/context.h
M /branches/citadel_context_rework
r8381 | davew | 2010-02-26 10:14:42 -0500 (Fri, 26 Feb 2010) | 2 lines
Changed paths:
M /branches/citadel_context_rework/citserver.c
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework
Oh great another great swathe of things to merge, lovely :-)
Fri Feb 26 2010 01:11:40 PM EST from dothebart
dotheb...@uncensored.citadel.org Subject: Citadel commit log: revision
8384
r8384 | dothebart |
r8385 | davew | 2010-02-26 14:17:15 -0500 (Fri, 26 Feb 2010) | 4 lines
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/context.h
M /branches/citadel_context_rework
r8386 | davew | 2010-02-26 14:53:01 -0500 (Fri, 26 Feb 2010) | 3 lines
Changed paths:
M /branches/citadel_context_rework/citserver.c
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework
r8387 | davew | 2010-02-26 18:49:09 -0500 (Fri, 26 Feb 2010) | 5 lines
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/sysdep.c
M /branches/citadel_context_rework
sunday. Messages made it
from one box to another after that...
does it still loop for you?
Mi Feb 24 2010 14:31:00 EST von davew @ Uncensored Betreff: async
stuff
More info.
The loop of error messages from Dothebart's IO code is not directly
related. Seems Citadel chat
Interesting.
That would explain a lot.
Thu Feb 25 2010 01:02:24 EST from IGnatius T Foobar @ Uncensored
Subject: Re: async stuff
Ok, it seems that the buffered I/O *is* causing a problem.
sysdep.c : 1147
HaveMoreLinesWaiting() *always* returns 1, until the session ends. As a
r8375 | davew | 2010-02-25 08:19:11 -0500 (Thu, 25 Feb 2010) | 1 line
Changed paths:
M /trunk/citadel/sysdep.c
Fix HaveMoreLinesWaiting. Fixes IG's problem with async.
.
Hey, I'll bet this even fixes the loop i've been seeing.
Thu Feb 25 2010 08:19:14 EST from davew da...@uncensored.citadel.org
Subject: Citadel commit log: revision 8375
r8375 | davew | 2010-02-25 08:19:11 -0500
?
Probably valgrind will easily tell you the exact location on first try.
Do Feb 25 2010 08:23:05 EST von davew @ Uncensored Betreff: Citadel
commit log: revision 8375
IG, this fixes the fundamental problem with async but there is something
else wrong though not related.
When webcit sends
I have a question regarding how (or rather when) Citadel limits the number of
sessions / contexts / connections.
At the moment the session limit is enforced by the protocol. In other words,
the connection is accepted, a context created and initialised, all the
associated events triggered and
r8376 | davew | 2010-02-25 11:07:11 -0500 (Thu, 25 Feb 2010) | 1 line
Changed paths:
M /branches/citadel_context_rework/citserver.c
M /branches/citadel_context_rework/context.c
Interrim commit.
I've taken a quick look at this and it may be me that broke it.
The fix that prevented crashing due to contexts not being ready and more than
one context with the same FD etc. You know the one thats really difficult to
explain.
It probably broke asynch operation.
I made several
I'm not sure you fixed this.
My earlier post indicates that a similar situation still exists.
Sun Feb 21 2010 14:37:29 EST from dothebart
dotheb...@uncensored.citadel.org Subject: Citadel commit log: revision
8354
Got it.
Yep, its broken, definately, Now I have that description its obvious.
To fix the problem with re-used fd-s before contexts got cleaned / removed
and crashing and all that I went to great pains to ensure we only chose
contexts with activity on the fd to bind to. I didn't consider that
IG, I'm trying to test this but my pidgin won't connect.
It reports unknown error.
In the citserver log I get Ignoring unknown tag bind just before the
connection gets closed.
Wed Feb 24 2010 08:33:14 EST from IGnatius T Foobar @ Uncensored
Subject: Re: asynch not working
r8370 | davew | 2010-02-24 11:32:08 -0500 (Wed, 24 Feb 2010) | 1 line
Changed paths:
M /trunk/citadel/sysdep.c
Fix async messages (hopefully).
r8372 | davew | 2010-02-24 13:26:10 -0500 (Wed, 24 Feb 2010) | 1 line
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/housekeeping.c
M /branches/citadel_context_rework
OK, so what exactly (versions) do I need to test this.
I can't get my pidgin to connect.
Wed Feb 24 2010 12:49:35 EST from IGnatius T Foobar @ Uncensored
Subject: Re: Citadel commit log: revision 8370
Fix async messages (hopefully).
Nope.
If you send a message from Cit client to pidgin the message doesn't arrive.
Got that bit.
Question. Does the message from Cit client arrive when pidgin sends a new
message to Cit client?
Wed Feb 24 2010 08:33:14 EST from IGnatius T Foobar @ Uncensored
Subject: Re: asynch not
IG, I've got my pidgin to connect now but I have a totally different problem.
As soon as my pidgin sends to the citserver the citserver dives off into an
endless loop in dothebarts new buffered IO code.
Now thats a different problem that causes me issues instead.
there is something more else where.
Wed Feb 24 2010 14:16:52 EST from davew @ Uncensored Subject: async
stuff
IG, I've got my pidgin to connect now but I have a totally different
problem.
As soon as my pidgin sends to the citserver the citserver dives off into an
endless loop
Ok, so what is filling my temp directory with file named foolog*
r8367 | davew | 2010-02-23 13:06:56 -0500 (Tue, 23 Feb 2010) | 1 line
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/context.h
M /branches/citadel_context_rework
r8368 | davew | 2010-02-23 13:12:54 -0500 (Tue, 23 Feb 2010) | 1 line
Changed paths:
M /branches/citadel_context_rework/include/ctdl_module.h
D /branches/citadel_context_rework/modules/jabber
A /branches
Would you believe it?
Tortoise SVN, the loveable SVN client for Windoze converts text files to \r\n
terminated during the transfer.
Maybe its just me but I think that is SO VERY WRONG. Surely something like
that should always move the file unmodified.
Go for it. Theres only one good thing about the current time stamps and that
is it tells me how far behind the yanks are :-)
Sun Feb 21 2010 17:45:16 EST from dothebart @ Uncensored
http://www.onlineaspect.com/2007/06/08/auto-detect-a-time-zone-with-javascrip
t/
hm... maybe
r8358 | davew | 2010-02-22 12:06:16 -0500 (Mon, 22 Feb 2010) | 1 line
Changed paths:
M /branches/citadel_context_rework/internet_addressing.c
M /branches/citadel_context_rework/modules/clamav/serv_virus.c
M /branches
r8344 | davew | 2010-02-21 08:04:26 -0500 (Sun, 21 Feb 2010) | 1 line
Changed paths:
M /branches/citadel_context_rework/citserver.c
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework
r8346 | davew | 2010-02-21 10:58:02 -0500 (Sun, 21 Feb 2010) | 1 line
Changed paths:
M /branches/citadel_context_rework/clientsocket.c
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework
r8318 | davew | 2010-02-16 09:19:56 -0500 (Tue, 16 Feb 2010) | 1 line
Changed paths:
M /branches/citadel_context_rework/clientsocket.c
M /branches/citadel_context_rework/clientsocket.h
M /branches
If I'm getting my head round what you are trying to do and I think I am then
your problem here is the same thing as was causing the crashes, that is, a
thread is in select with an FD_SET containing the FD before you made it be
-1. What we need to do is interrupt the select so that every thread is
Thats why I didn't impliment the signal to awaken things in select()
In fariness though there are not that many places where it would be needed.
Almost nothing blocks so it already checks to see if the return code is
useful and reacts appropriately. If the syscall fails the failure action is
IG,
Have you run the current HEAD on that machine with lots of spam?
I made a change that might have plugged the hole properly. If it still booms
then I think we should leave it untill I get the context rework branch
sorted, merge that in and fix it that way.
Damn, I realy thought that was the hole.
It has to be something like that but obviously its not that particular hole.
Hmm, do I want one of my small networks that is connected via ADSL at 2M/512K
to be battered to death with spam when you change the MX
Hmm, not sure about that one.
r8265 | davew | 2010-01-31 12:15:06 -0500 (Sun, 31 Jan 2010) | 2 lines
Changed paths:
M /trunk/citadel/sysdep.c
Another attempt to close the hole causing crashes when CC==NULL
Sun Jan 31 2010 09:10:58 AM EST from dothebart @ Uncensored Subject:
Re: Citadel commit log: revision 8262
doesn't mstone offer a way to produce similar effects when ran from several
hosts in your local network?
It does.
Problem is I don't have several machines
Hi guys,
I think I have a reason for this and a solution.
I'm going to try it here first in case there are side effects.
We all need to be aware that my test boxes almost never exibit this problem
and the one box I have observed it on was a one off only. Basically it is
almost impossible
r8262 | davew | 2010-01-30 11:11:50 -0500 (Sat, 30 Jan 2010) | 6 lines
Changed paths:
M /trunk/citadel/sysdep.c
Attempt to fix the crash that results from a new sessions context
not existing before the command loop
Hopefully that cures the CC == NULL problem.
As I said before I can rarely reproduce this problem and so had to resort to
mentaly tracking thread re-schedules and the like.
This fix closes a hole that was already present but has recently been made
bigger.
The theory is that a context dies
Yep, that looks like exactly the bug. Must be something not atomic somewhere.
I've done with work for this week so I'll probably get to it this weekend :-)
Are we now back in full development cycle?
In other words is the trunk HEAD now unstable and ready to be played with?
Thu Jan 21 2010 03:31:16 PM EST from davew @ Uncensored Subject: Re:
Citadel commit log: revision 8249
Wed Jan 20 2010 12:39:12 PM EST from IGnatius T Foobar @ Uncensored
Subject: Re: Citadel commit log: revision 8249
For what it's worth, this issue is getting exposed
Seems I was confused again.
The fix is in and should be good but I want to check a few things.
Those of us that have seen this bug is it on a 32 bit machine or a 64 bit
machine??
I think I've seen it on a 64bit machine but can not confirm at the moment
Definately NOT seen it on a 32
Wed Jan 20 2010 10:53:09 AM EST from dothebart @ Uncensored Subject:
Citadel commit log: revision 8249
Mi Jan 20 2010 10:50:48 EST von ajc a...@uncensored.citadel.org
Betreff: Citadel commit log: revision 8249
Fri Jan 15 2010 04:50:01 AM EST from Mathew McBride @ comalies
Would there be an easy way to determine if an inbound mail message has been
moved away from Mail by sieve?
I would like to change the notify module so that I does not invoke a
notification in the above case.
Tue Jan 05 2010 06:33:23 PM EST from dothebart @ Uncensored Subject:
Citadel commit log: revision 8206
quiet a bunch of warnings...
I guess the wiki one could be problematic...
I'm actually running that code on both 64 bit and 32 bit. No problems yet but
it
Wed Jan 06 2010 04:51:00 PM EST from ajc a...@uncensored.citadel.org
Subject: Citadel commit log: revision 8209
r8209 | ajc | 2010-01-06 16:50:59 -0500 (Wed, 06 Jan 2010) | 2 lines
Changed paths:
M
Thu Jan 07 2010 04:15:49 PM EST from dothebart @ Uncensored
dave, please have a look at these cores over here on uncensored.
though they're from xmpp and imap, they've got one thing in common:
session_specific_data = 0x0
causes the crash. it shouldn't enter the commandloop
r8211 | davew | 2010-01-10 07:41:01 -0500 (Sun, 10 Jan 2010) | 8 lines
Changed paths:
M /trunk/citadel/context.h
M /trunk/citadel/sysdep.c
Fix race condition that caused segfaults in imap and xmpp as seen on
uncensored
r8212 | davew | 2010-01-10 10:02:32 -0500 (Sun, 10 Jan 2010) | 4 lines
Changed paths:
M /branches/citadel_context_rework/context.c
Change to ForEachContext() now the CC is preserved so that it is correct
for each job
What tools are people using to access Subversion?
Are you using the command line or some front end tool.
I'm trying to find something that will make it easier for me to keep track of
changes in the trunk and merge them into my branch.
r8219 | davew | 2010-01-10 16:45:24 -0500 (Sun, 10 Jan 2010) | 2 lines
Changed paths:
M /branches/citadel_context_rework/context.c
code clean up.
r8222 | davew | 2010-01-10 18:49:00 -0500 (Sun, 10 Jan 2010) | 5 lines
Changed paths:
M /branches/citadel_context_rework/citserver.c
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework
r8196 | davew | 2010-01-04 08:50:03 -0500 (Mon, 04 Jan 2010) | 2 lines
Changed paths:
M /branches/citadel_context_rework/context.c
M /branches/citadel_context_rework/context.h
M /branches/citadel_context_rework
Sun Jan 03 2010 03:51:30 PM EST from dothebart @ Uncensored Subject:
encoding problems
ok, the problem is:
mime_parser.c line 396:
extract_key(charset, content_type, content_type_len,
HKEY(charset=));
we search: 'charset='
that mail has 'charset ='
r8192 | davew | 2010-01-03 17:01:56 -0500 (Sun, 03 Jan 2010) | 2 lines
Changed paths:
M /branches/citadel_context_rework/citadel.h
M /branches/citadel_context_rework/configure.ac
M /branches/citadel_context_rework
OK, who broke it :-)
Seems that webcit on the SSL port is broken 8080 port is still working.
Mon Dec 28 2009 08:08:41 AM EST from davew @ Uncensored Subject:
Uncensored broken via webcit HTTPS
OK, who broke it :-)
Seems that webcit on the SSL port is broken 8080 port is still working.
Oh and perhaps I should have said it seems to be a Webcit problem
Thu Dec 17 2009 08:01:42 PM EST from IGnatius T Foobar @ Uncensored
Subject: Re: Citadel commit log: revision 8139
Sounds very cool, but is this going to be one of those things that only
davew understands and therefore becomes difficult to maintain?
Take a look
Thu Dec 17 2009 09:28:59 PM EST from dothebart @ Uncensored Subject:
7.70 release
Do Dez 17 2009 14:58:43 EST von davew @ Uncensored Betreff: 7.70
release
We can each have a branch to work on different features and are each
responsible for keeping our branch in line
Fri Dec 18 2009 02:03:32 PM EST from dothebart @ Uncensored Subject:
7.70 release
Git uses UIDs instead of counting up a number; Which makes a source
commit over several branches uniq.
Git also knows to pull over changes from other branches in a relative smart
way; so you can
r8141 | davew | 2009-12-18 14:50:26 -0500 (Fri, 18 Dec 2009) | 5 lines
Changed paths:
M /branches/citadel_context_rework/server.h
M /branches/citadel_context_rework/threads.c
Finer grained locking of worker load
r8142 | davew | 2009-12-18 15:02:19 -0500 (Fri, 18 Dec 2009) | 3 lines
Changed paths:
M /branches/citadel_context_rework/context.c
Fixed a warning and updated a comment to better explain
how ForEachContext() handles
r8144 | davew | 2009-12-18 15:24:14 -0500 (Fri, 18 Dec 2009) | 3 lines
Changed paths:
M /branches/citadel_context_rework/include/ctdl_module.h
M /branches/citadel_context_rework/msgbase.c
M /branches
r8146 | davew | 2009-12-18 15:42:20 -0500 (Fri, 18 Dec 2009) | 3 lines
Changed paths:
M /branches/citadel_context_rework/modules/crypto/serv_crypto.c
Silence 3 warnings by adding (unsigned char *) to argument 4
r8147 | davew | 2009-12-18 15:47:08 -0500 (Fri, 18 Dec 2009) | 2 lines
Changed paths:
M /branches/citadel_context_rework/modules/newuser/serv_newuser.c
Silence a warning about ambiguous ! in an if()
Fri Dec 18 2009 02:43:06 PM EST from dothebart @ Uncensored Subject:
7.70 release
Fr Dez 18 2009 14:16:11 EST von davew @ Uncensored Betreff: 7.70
release
I've lloked at it too, even read some articles and how-to's.
At first glance its quite complicated
Fri Dec 18 2009 05:52:58 PM EST from IGnatius T Foobar @ Uncensored
Subject: Re: Citadel commit log: revision 8139
TDAP purges a message when its reference count reaches zero. Dig dis:
* Whenever a message is saved to a room, or removed from a room, we write a
record to
Sat Dec 19 2009 04:57:29 AM EST from dothebart @ Uncensored Subject:
7.70 release
Fr Dez 18 2009 15:52:16 EST von davew @ Uncensored Betreff: 7.70
release
Hang on are you using SVN or GIT?
Or are you saying that GIT can pull from an SVN and push back into an SVN
r8129 | davew | 2009-12-17 10:58:25 -0500 (Thu, 17 Dec 2009) | 4 lines
Changed paths:
M /branches/citadel_context_rework/server.h
Added #define for new event type EVT_IDLE
EVT_IDLE will be sent when citserver
r8130 | davew | 2009-12-17 10:59:55 -0500 (Thu, 17 Dec 2009) | 2 lines
Changed paths:
M /branches/citadel_context_rework/threads.h
Clean up some commented out bits
r8131 | davew | 2009-12-17 11:01:57 -0500 (Thu, 17 Dec 2009) | 3 lines
Changed paths:
M /branches/citadel_context_rework/user_ops.c
Some minor speedups by moving some extract_foo items outside of
critical sections.
r8132 | davew | 2009-12-17 11:08:06 -0500 (Thu, 17 Dec 2009) | 7 lines
Changed paths:
M /branches/citadel_context_rework/scripts/mk_module_init.sh
M /branches/citadel_context_rework/threads.c
Added a THRD command
r8133 | davew | 2009-12-17 12:41:57 -0500 (Thu, 17 Dec 2009) | 3 lines
Changed paths:
M /branches/citadel_context_rework/citserver.c
Use the functions for getting load averages in the info command
so that locks
1 - 100 of 508 matches
Mail list logo