I think that fixes OpenID. There was a problem with the conversion to
strbuf in that code.
I do like the StrBuf class, but we seem to incur a lot of bugs during
conversion. That being the case, I would like to decide right now that we
should *not* start converting anything in citserver
Question: how do you express a conditional of if this variable is longer
than zero bytes in the template engine?
/me is avoiding #citadel on freenode until email kindergarten is over.
Ok, I see how that works. I added another one for the 'Subject:' line.
What I want to do is test whether a message contains a recipient, and
output or not output the word to in the headers.
Uncensored is now running on svn head as of right now. Beta testing has
begun. All further commits should be strictly bugfixes until we get this
next release out the door.
It was just pointed out to me that the Delete button is missing from
the mailbox view. I hadn't noticed because I always use the delete key.
Can we get it back?
What does it do differently when you register it as AJAX, anyway?
ok, i've tried it with cadaver, its working the way its now.
i wasn't able to connect groupdav clients... tbird + sogo didn't work at
all
(js-errors dead...) and I wasn't able to find out how Kontact is
configured
these days. sick.
There may be some problems with
this brinngs the nice background into the baby obama message. I'm
wondring
how we could transport whats between head and body into our display?
probably theres some CSS information in there...
There's definitely some important information in there. Notice that the
body style
Doing some more GroupDAV troubleshooting...
When the GroupDAV module calls locate_message_by_uid(), it already has
the wrong filename in DavPathname by the time groupdav_main() is called.
Hmm... this looks bad:
groupdav_main.c: TODO
Maybe span ?
Ok, I've updated the Easy Install copy of webcit from the
webcit-738-stable branch.
If this doesn't appear to be introducing any new bugs then we can release
this.
My reply seems to become interpreted older than the original post? How
does
the RSS Reader create the timestamp?
It parses the date/time field from the feed, if it's there, and if it's
readable. Otherwise it uses the date that the feed was parsed.
Again, no unpleasant surprises here; we are in feature freeze and there's
no way I was going to touch Citadel with this. Besides, we still need a
couple of releases between the deprecation of the useless populate LDAP
module and its replacement with a useful authenticate from LDAP module.
All of your issues are taken care of? How about you, Matt? On the server
side I'm comfortable with where we are right now.
Has everything been tested in IE6, IE7, and Safari/Opera/Chrome/etc ??
Go for it. It would be a good sell in MS environments.
That's part of the goal. I recently spoke to someone who had done some
independent research about Citadel and he told me that when people
discuss Citadel it's generally positive, but the one thing that they
complain about is the
table class=mailbox_summary id=summary_headers cellspacing=0
style=width:100%;-moz-user-select:none;
is definitely missing quotes / having wrong quotes. where do they need
to go
at?
There should be a quote after the zero ... cellspacing=0
Ig, if yu rebuild the msgformart files, please use './buildpackages
i18n' pay
attention that webcit will find a fresh copy of the templates in its
static
directory. Please recheck that there realy are strings inside of the
generated 'c' file.
Ok, so does this mean that if
i'd like to release beta 7.42 with that. any objections?
Go for it!
Oh wait ... I'm reading RFC 5464 a little more closely now ... and it
seems that they've actually *removed* the stupid part with which we were
headed towards incompatibility. Up until Draft 12 it contained a *really*
*ugly* LIST-EXTENDED command that augmented the LIST and LSUB commands
to
We have discovered a new template problem.
The previous behavior was for Reply and ReplyQuoted buttons to appear
in the Bulletin Board view, and for Reply and ReplyAll buttons to
appear in the Email view.
(When using the Email view, *all* replies were Quoted.)
In the
I just noticed that user names in the webcit wholist (both the big one and
the iconbar one) are now appearing urlencoded. Did something change?
so that nice link to citadel.org in every login screen won't be
evaluated at
all...
shouldn't we allow that in a way? is there a way?
For now, it's fine to just tell robots to go away.
Citadel v8 will need something more sophisticated than that, because the
new
Because, y'know, comments that are DFSG-compliant are better than comments
that are guaranteed to be accurate to the RFC. Grumble grumble grumble.
Hey you think we can apply this to WebCit to get it to recognize primary
email address when editing vCards made with Sync Kolab?
Hang tight, I'll do this the right way
7.43 (another beta release) is now committed to svn, and is available on
Queasy Install, and is also running on Uncensored.
Copy isn't always appropriate either ... although I suppose a copy from a
public room into a mailbox is. Now isn't the time to be adding features,
though.
After a lot of review and consideration, I'm just not happy with the extra
complexity introduced by the xmacros stuff. I really think I'd prefer to
keep it simple, and if serv_vandelay is being replaced, now's the time to
make that decision, before any more time is spent on it.
It's an
I am simply not planning to write a DTD, that's all. The program's output
looks like this:
?xml version=1.0 encoding=UTF-8 ?
citadel_migrate_data
version743/version
config
c_nodenametesthost015/c_nodename
c_fqdntesthost015..com/c_fqdn
c_humannodeFoo Bar and
The c_ prefix isn't for citadel it's for config. I don't care about
the tag names being globally unique, as long as they're unique within this
one application. All of our configuration variable names start with c_
already, and wherever possible I'm keeping the tag names identical to the
I'm using ctdlphp to write a blog front-end that exposes public folders
and
shows messages that start a thread as blog posts (full html with
embedded
images etc) and threaded replies as blog replies.
Sounds *very* useful. We don't currently render threads but the
underlying
Ok, well, if this cannot be made to work reliably and immediately, then
these changes should be rolled back.
If we want to do SSL-encrypted Jabber, here's how to do it:
1. Edit serv_xmpp.c and search for starttls
2. Change #ifdef HAVE_OPENSSL__COMMENTED_OUT to #ifdef
HAVE_OPENSSL
3. Note that we have the exact same problem that is experienced with the
port-5223 implementation: the
v7.44 is now in SVN and is also available via Queasy Install.
With this update we move from Feature Freeze to Code Freeze. Target date
for final release is early April.
The normal exit path is through master_cleanup() in citserver.c, line 231
if it's an error-free shutdown. exit(exitcode) where exitcode == 0.
Yikes. DLAT is missing the MIME metadata. Must fix that immediately.
Ok, there it is ... you should be able to switch back to DLAT now.
As per my comments on irc earlier today ...
* I really think it's a bad idea to throw away 30 years of unix tradition
by declaring some value other than 0 to be the non-failure exit code.
This will certainly mess up a lot of third-party scripts out there.
* If we have the need for
Ok folks, please observe that we now have a STABLE branch. Any
last-minute bug fixes made between now and the 7.50 release need to be
committed to *both* the stable branch and the trunk.
That doesn't mean we can be careless with the trunk, though. I would
like the 7.6X development
Was anything changed in the conditionals other than the numbers? (I just
need to know how extensively we need to re-test.)
Ok, that about wraps it up for LDAP authentication. If anyone wants to
test it out I'd love to hear what you think of it.
I ended up having to offer two different modes, one for the standard
RFC2307 schema, and a different one for Active Directory's ugly
nonstandard schema.
Error building webcit (trunk) --
make: *** No rule to make target `ical_maps.o', needed by `webcit'.
Stop.
I think I know how we can avoid fetching each message twice during a
calendar render. We've already got MSGP and MSG4 ... and we *already*
prefer calendar items:
serv_puts(MSGP text/calendar|text/html|text/plain);
So it would seem that if we're using MSG4 to read the contents
I'm suggesting that a search for an empty header (wefw|) should also
match
messages without that header.
Ok, this seems reasonable enough that you should go ahead and try it;
please watch for side effects. A good place to do regression testing
would be load/save of preferences
Ok, I've located the source of the problem where you can't log in as a
user with spaces in the name.
striplt() doesn't just strip leading and trailing whitespace anymore.
The current version now seems to strip *all* whitespace.
This is probably causing other issues as well.
*does bruce lee sounds*
I'd be more impressed if you emulated Jimmy Superfly Snuka :)
Egads, that was 26 years ago. Now I feel *really* old.
I can't seem to build WebCit:
vcard_edit.o: In function `display_vcard_photo_img':
/root/ctdl/trunk/webcit/vcard_edit.c:1207: undefined reference to
`VCardLoad'
vcard_edit.o: In function `do_edit_vcard':
/root/ctdl/trunk/webcit/vcard_edit.c:796: undefined reference to
`VCardLoad'
*smack*
That did it. I was looking for VCardLoad() in the webcit sources.
samjam: this is all great stuff. I wasn't familiar with socat; it looks
like just the thing to make the php framework usable without the clumsy
session proxy.
Ok, looking good here. I'll have to learn how to use all these new toys
when I start building some of the new components.
I did notice something wrong, though: the Notes view is now broken.
Instead of showing the notes themselves it now shows MIME headers.
I'm hoping to get some
Comments? Complaints?
How many weeks did we just add to the 7.60 development cycle by doing
this now?
That sounds expensive.
The autopurger lives in modules/expire/serv_expire.c
Do you want notifications of a message's removal from disk, or from a
particular room?
If it is the latter then I suppose we could write out a flat file as we
do the purge, and process it
For a manual delete, we remove the message number from the room's msglist
and then write a record to refcount_adjustments.dat containing the msgnum
and the number -1 to indicate that the message has been removed from one
room.
When the autopurger runs, one of the last things it does is
Thanks for the patch; it has been applied.
We don't really know what happened to davew. He just disappeared one day
and never came back.
Ok, so we have our first batch of messages in a mailing list room that are
undeliverable since the improved mailing list code went online. I've got
a bunch of bounces in the Aide room, which is exactly where they're
supposed to go.
So ... hopefully everyone is *not* getting bounces in
Ok folks, as of today we are freezing for the 7.60 release. No new
features. Bug fixes only.
Heh - I didn't mean hurry up and do it I meant, any ideas on how it
could
be portayed?
Rather simply, actually. Amend the output of the RWHO server command to
include an extra field indicating an IM-capable session. Any session not
advertising the ability to receive IM's should
Ok, here we go. I'm upgrading Uncensored to svn trunk tonight. It's beta
time.
Well, that didn't work out too well. I've reverted back to svn trunk
citserver with stable-75x webcit.
Problems so far:
1. No graphics. For example, the iconbar is text only.
2. Sometimes when I log in it takes me to a room which does not exist. The
room's name is some random number.
Citadel's RSS reader is a hand-coded mapping of RSS directly into the
Citadel message base. It's fewer than 600 LoC because we *didn't* bolt on
some big library that has lots of features we'll never use.
(Unfortunately the same cannot be said for the HTTP-client portion, but
libCurl did
Please note that the bug fixed in r7607 was not present in stable.
I do notice we have a new bug, though. When you hit Reply to an email
that was sent from a local user to another local user, it isn't quoted. I'm
guessing that this bug appeared during the introduction of Save to Drafts
-- a
This is the third time I've made that fix. Let's try not to break it again.
Hey, what was that problem we were having where HTTPS webcit logins were
failing? I saw it at home tonight on a computer that had the HTTPS webcit
for Uncensored bookmarked as its home page. Weird thing, it seems to just
say your password was not accepted and goes back to the login page, but it
This was the only thing that was really bugging me before we could release
7.60. Are there any glaring problems that need addressing, or can we get
the release process going now?
I agree that edit is confusing in most vires, but most mailers to
provide
the option to edit a message as a new message, so I'm also not entirely
comfortable removing it.
I'm not comfortable having it there unless it works! I'm sure it broke
on its way into the tree, but
Does the Citadel project have any start on an automated test suite? Â
Or
any documentation of manual tests, for that matter?
Yes, look at stress.c for a small test suite. It isn't comprehensive but
it does work.
Yes, all good ideas; unfortunately it's one of those things that the core
team just doesn't have enough time available to do. We've definitely left
the door open for this to be done in the future ... or in the present if
someone volunteers to help out :)
Looking into the group select bug ... so far I'm seeing that these
expressions are failing:
markedRowId = parent.ctdlRowId;
var rowId = parent.ctdlRowId;
The expression parent.ctdlRowId is undefined. This causes the group
select operation to mark rows 0 through 0, which
Ok, there are still some problems in this. I don't think the data model
has been fully updated. There are still references to *.ctdlRowId in the
code. I updated one of them, and that produced valid starting and ending
row numbers, but the group select still didn't work.
I also see a
Still trying to make sense of this ... what I'm starting to see now is
that it attempts to determine the range of messages for group select with
row numbers based on the order in which they were received from the
server, NOT the order in which they appear on the screen. So, if the
table was
Here's what I've done so far:
* Renamed rowId and markedRowId to rowIndex and markedRowIndex, to
reflect the fact that we are working with current on-screen index
positions rather than the order in which the rows were loaded from the
server
* Used the expression parent.rowIndex to
Aha! Now this is some real progress:
WCLog('startMarkingFrom=' + startMarkingFrom + ', finish=' + finish);
for(var x = startMarkingFrom; xfinish; x++) {
WCLog(Marking row + x);
markRow(parent.parentNode.rows[x]);
}
Still a little wonky but it's
Ok, all yours, Matt ... here's where we stand:
* Group select is working
* Multi select is partially working. Clicking on an unselected message
will select it, but clicking on a selected message will not unselect it.
* Hitting delete appears to perform the correct server
It's an edge case, to be sure, but we might as well fix it now before
someone inevitably finds it later and starts complaining about it.
I was skeptical about the json format and client-side rendering at first,
but now that I'm getting used to it I'm starting to see how versatile it
is.
And it looks like they've already marked it as a dupe.
Evolution's failure to behave properly with namespaces is a colossal
failure on their part. They should either map the namespaces properly or
not support them at all.
Of course, this is Ximian (aka Novell) we're talking about
Ok folks ... Citadel 7.60 is now in soft launch. We'll announce it
soon.
Got these errors when I tried to upgrade webcit (clean build, and I had
already upgraded libcitadel)
tcp_sockets.c: In function bStrBuf_ServGetBLOBb:
tcp_sockets.c:217: error: bErrb undeclared (first use in this function)
tcp_sockets.c:217: error: (Each undeclared identifier is
We've fixed enough heinous bugs over the last couple of days to justify a
7.61 release, so I went ahead and did it. This was worth doing even if
7.62 comes right behind it.
So it seems that our biggest bug right now is that people still are having
trouble logging in using WebCit and SSL. This would seem to imply that
the buffered I/O thing is still broken?
It only happens with certain browsers.
Ok, for those of you who were looking for a convenient way to auto-submit
Citadel messages into SpamAssassin for training its Bayesian filter,
here's a quick and dirty utility to do it.
READ THE README FILE. READ IT SEVERAL TIMES.
In this case, TFM describes in great detail how I feel about people who whine
for support on software that has been declared as unsupported. It's a good
read. :)
Is anything currently in the trunk unstable? Or can we just release
what's there now?
Hi -
Hi! :)
No need to retro-patch; I'll release it as 7.64.
Ok, but we have to be REALLY REALLY CAREFUL here. citserver will not
tolerate the kind of subtle bugs that have crept into WebCit during the
adoption of the StrBuf class.
The other issue is performance. That function was *highly* tuned, after I
took a lot of abuse for it running very slowly when operations are
performed on very large mailboxes (2 messages or more). If the
introduction of StrBuf incurs a performance penalty, this is a problem.
* fix it some, but not all.
As of right now, the bug is still there -- it's still looping endlessly
when I try to open the inbox on my dev system.
I just upgraded citadel and webcit to the newest versions and noticed
that on
my summary page none of my calendar info is showing up.
Anyone else see this?
This sounds like something else that got broken when the read loop was
rearchitected?
And why do you guys keep changing this stuff but not testing it to see
if things work... I think for about two versions it would not even
save a sieve script for whatever reason, and now that it does again, the
sieve script gets cut off upon attempts to save the
On the good news side, however, citserver is no longer looping on
CtdlSetSeen().
Ok, that takes care of the truncate issue, but there's still something
weird about this. Manipulating the flag on the very last message in the
list doesn't work. I see the code there to handle that condition, but I
don't see why it isn't working.
I like the idea of unit tests, but I don't understand how they can be
applied to code which implements a user interface -- which happens to be
where the bulk of our bugs are right now. How is that normally achieved?
(The discussion on unit tests has been moved to the Programming room)
I don't know. They came back all by themselves!
Funny, I did two commits, one of them got its log message through, the other
didn't.
Anyway, I re-did the logs because they were looking pretty bad. Now it's in
pretty HTML with the usernames in bold and each instant message wrapped up in
p/p tags. I also set the message author to the name of
..and that's why the double messages didn't show up during testing: I
only had the two users logged in one session each.
On most of my desktops I have my Citadel client set up so that if it sees
I have a Jabber session open to the server, it throws away any incoming
instant messages.
That's good to know, but the double messages to which I was referring
aren't the ones originating from mycastlerealm. We were actually tracking
down a problem with the new instant message logs, in which we now save
entire conversations instead of one instant message at a time. If the
701 - 800 of 1678 matches
Mail list logo