Ok ... Citadel 7.24 is out, and tonight I'm going to do a release of
WebCit 7.24 with the last couple of bugfixes. After that, I
am going to commit a big patch that moves all of mime_parser.c and all of
tools.c into libcitadel. Now is the time to start checking out and
building libcitadel.
There it is, folks. Go to town.
wtf?!?!? I put in a change last night to fix mail that was getting
posted to an address book room.
Ok, I probably killed my citserver while it was indexing and then got
impatient and didn't wait for the database to flush. I'll keep an eye on
it.
I don't know much about the build system except that it's supposed to
generate the dependencies automatically.
Yeesh. My database is badly corrupted. But I think it's my fault, not
the software's.
Read messages with the text client.
I fixed that on Friday evening.
Commit log says it all, really. Move the decoding intelligence into the
client.
I'm using the Makefile from libspawner as a scaffold for building
libcitadel. Let's see how portable it ends up being.
There it is. We should test with a bunch of different calendars from
icalworld.com ... it's a good source of foreign events.
It also demonstrates how badly we need to profile the calendar code.
Showing thousands of events is quite slow.
Again, we should avoid putting too much effort into Notes because the
entire thing needs to be rewritten -- not just the UI, but the
underlying functionality. Our notes are text. They should be in VNOTE
format, like this:
BEGIN:vnote
VERSION:1.1
PRODID://Bynari Insight Connector
I continue to test webcit with IE browsers. that seems to work
correctly with the recent commits.
Yes, I see things working that weren't working before. Nice work.
One thing that still seems to be missing: the address book popup doesn't
appear when you click Contacts while
Has any one else noticed the new files counter that tells you how
many files there are in a directory room?
I saw that just now, it's a nice thing to have, but I do have several
concerns:
1. How much more server activity does it incur? Has this been profiled?
2. If you go
Actually I'm checking for QR_VISDIR. If the directory isn't visible
Even better. But if a directory room is made into a non-directory room,
QR_VISDIR might not be un-set... so you might want to check for *both*
bits.
FYI: Easy Install is now deploying Expat 2.0.1 in order to enable the new
RSS feed reader. Some sites that upgraded over the past couple of days
might have gotten the new functionality anyway if they picked up
/usr/lib/libexpat.so.
i ran into troubles with that non-byte-by-byte-read fix.
I'm looking at citadel_ipc.c again today.
If it would help to make the symptom less specific to a particular
function, here it is:
When a chunk of data is being read from the server, it is removing blank
lines.
Ok, now you can relax and work on this at your own pace. :)
What else is holding us up from our next release?
davew: a delete button is nice, but do keep in mind that I have eventual
plans to rip out the entire notes module and replace it with something
completely different.
There's more to it than that.
The big issue is multipart/alternative where we have to iterate through
all the MIME parts in order to determine which of the alternatives we want
to see, based on the client's declared preferred types.
There are all sorts of nasty variables, though,
We've got wiki rooms already working, and it's actually quite nice. It's
disabled because it isn't finished.
Right now I don't think it's a good use of our time to start changing
everything around. We have all of the resources in place, we just have to
make use of them.
The
davew: good point ... if the message does not arrive via SMTP, the
envelope information may be less than perfect. Perhaps that is where we
should be starting?
I'm also wondering whether we should turn off bugmail, or maybe move it
to another room.
Fixed! All this time I thought it was a webcit bug. Turns out it was a
bug in the MIME parser, which caused MSG4 to send WebCit an empty message.
Putting the MIME boundary as anything other than the very last key in the
Content-type: declaration is rare, but quite legal. Fixed now.
hmm... tabs.c completely broken under IE 6 ... only the first tab screen
shows, the rest are empty. Also the rounded corners are gone. I haven't
tested it until today, so I don't know whether I broke it with my code to
highlight the correct tab, or if the stylesheet or something is broken.
Hey folks, I think we should feature freeze until we get the new WebCit
released. Y'all are writing some great code but we need to stabilize and
get it out to the people.
Ok, two comments on that:
1. We are, in fact, providing correct envelope information to Sieve. It
is collected during the SMTP transaction, saved in hidden fields in the
Citadel message headers, and passed along to libSieve when asked.
2. I believe the reason for requiring the
uggghh... I spent a couple of days focusing on work and now I'm way behind
on Citadel stuff ... there's something like 60 new messages in Citadel
Support and I'm having trouble keeping up with it.
Could someone please summarize what I need to work on in order to get our
WebCit release
Actually it looks really good.
Ok, if you're going to do it then I won't try to do it myself. I can
definitely tell you that if we do any more tabbed dialogs, the API in
tabs.c will be used. The way we did it in Edit or delete this room was
kind of old. The tabs.c API is much better, as it is more DHTML based; no
page
I think we're getting close to thoe point where we can do a new WebCit
release?
You're right, this API is the best way to do tabbed dialogs. I have
to clean up my code and to use it for Advanced room commands menu
I would recommend that we *not* do that. The reason is because the tabs
in the Advanced Room Commands menu draw a lot of data from a lot of
Oooh, I forgot that I used tabs.c somewhere else too. It's used in any
room that has an address book view (such as Contacts). And it's already
got the new look. Cool!!
davew: you can share a private room, no problem.
extract_key.
I'd apreciate some help here, or, maybe Ig, could you have a look at
that? I realy don't know what exactly this should do, so its hard to
tear it appart and put it proper together again...
extract_key() parses MIME headers and extracts the key/value parameters
contained
does the output of LKR* tell me what the prefered mime type of that
room is? vcard, ical, rfc822...
Yes, the listing commands include both the default view and the user's
current view.
Is it possible to get some sort of form on the Wiki that people can
fill in with stats about their system or could we automate it by
creating a module that the admin can activate, creates a room and
logs stats daily. Then we could share the room back to uncensored and
we can see
Oct 12 2007 9:00am from fleeb @uncnsrd
Oct 12 2007 12:59am from IO ERROR @uncnsrd
Maybe the next version should be 2008.
*smacks IO*
*smacks IO again*
Back to you, fleeb.
davew: it seems that the issue you are experiencing with stray garbage
files in the top level directory isn't just local to you -- I'm getting it
too. And you were right about it being network related -- I am running
three Citadels on the same host, from the same server executable, and only
Ok, all done. dothebart: I hope I'm not stepping on your toes here, but
that code really needed simplification. It took a couple of hours to get
it right but we've got code now that makes only minimal use of time_t and
doesn't require storing the expanded date/time fields in the event list.
Ok, so it looks like webcit needs to wait for a calendar fix, but we'll
release v6.22 of citserver now that LDAP is working.
Feh. I just can't seem to get myself used to the fact that we're living
in the 7.xx world now. We probably ought to go 8.xx or something :)
I wiped the webcit tree and got fresh svn and its fixed. Now why
would that be???
I don't know, but it happened to me too.
Sure, we can do that.
The other issue with calendar items, that we've got to address soon, is
that we don't expand recurrences.
Have I mentioned how insanely great you guys are? :)
My fault. Again. Fixed now.
serv_pager.h is now added. Sorry.
oops.
My fault. I applied Matt's patch and still didn't get all the new files
in.
Should be fixed now.
I was of the impression that it was just for address book
entries.
The current implementation is just for address book entries. The idea is
that if you have an email client that can only use LDAP as its address
book, Citadel can populate it for you. Anything saved to the
I'm not terribly fond of Sieve either, but we chose it because it's the
standard. Sieve scripts can be written for other systems and moved over
to Citadel without any problem. There's value in that.
On the other hand, we've had numerous discussions, and even several
aborted attempts,
I don't understand, I commit a new styleshhet then some minor
change. I see nothing appearing in the developpment room.
Commit logs are not getting posted right now. I don't know why. I will
fix it soon.
It turns out that Versamail wasn't pulling anything because it gave
us the IMAP search terms in parenthesis, where it wasn't expected.
I've replaced the calls to strcasecmp with calls to strstr which will
pick up these cases.
Thanks for the patch. After reviewing it, I
ok, now that we seem to have another citadel release, can we focus
a bit to get out the next webcit? as it contains some tuning on
I've just pushed the latest SVN HEAD out to Uncensored, so we'll see if
that shakes out any bugs we haven't fixed yet. I'm going to write a
Excellent, I just noticed that you did that. Thank you!
I realize that it is not finished -- there are definitely a few places
where we have to make some changes. But it is definitely a MUCH better
look than the previous theme.
(If anyone wants to see it ... we still have
hm, did i mis-operate tcpdump? maybe forget -s0 :( i'll redo it.
No, I think you've got something else wrong on your server. It looks
like you're actually sending truncated commands to Citadel.
can we patch that on the tarballs and easy install too?
Done. World, meet Citadel 7.21.
Ok, I have built a source tarball of Citadel 7.20. Package builders, now
is probably a good time to do your thing.
Ok, I have built a source tarball of Citadel 7.20. Package builders,
now is probably a good time to do your thing.
Oh, and please use r5473 for packages, because I'm already making changes
that shouldn't go in the packages.
Mailing lists customarily have [LISTNAME] in the subject.
Ok folks, we have the latest SVN HEAD on Easy Install now. Since we can
push updates to Easy Install very quickly, we have the ability to correct
problems quickly. If everything goes well over the next couple of days we
can do tarballs and packages as well.
davew: since it hasn't broken anything yet, let's run with it and be
prepared to address the issue if and when we actually start to see what
the failure modes (if any) look like.
Functions like stresc() were written with readability in mind rather than
performance. This made sense at the time, but sure, let's go ahead and
optimize now. Just be VERY careful not to introduce bugs. I'll test now.
davew: I don't think it'll break networking. My only concern is that
it'll display an incorrect address in some circumstances. Networking
itself doesn't perform an RFC822 output.
Further thoughts on this...
I wonder if libSieve would be content with a less than fully qualified
domain name. You could try appending the Citadel node name -- if the
message is en route to another node, you'll find that node name in the D
(Destination node) field, otherwise use the
I think it had something to do with Internet addresses showing up in
places where they shouldn't. It was late June when I tested it. Think
think think...
thierry: we did a WebCit release recently, so don't worry about it. I
just want to get a new Citadel server out soon.
thierry: the new look is beautiful.
* The extra rounding on the corners looks good
* The colors are VERY nice
* I really like what you did with the buttons. They blend in nicely.
On the other hand, I'm not sure about the tabs on the left side of the
admin
I have tested this fix as thoroughly as I can given I'm not sure all
the possible circumstances for calling OutputPreLoadedMsg()
Obviously I don't want to committ it till we know its ok.
Hmm ... I considered that but then realized that it would break something
else.
What happened to the stabilize and release?
Good question. I think we should focus on that right now; it's been
quite a while since our last release of the Citadel server.
Let's try to get our house in order so we can give the world a new server
release before embarking on
Fix to allow display of tasks created using Bynari connector from
Outlook.
davew: how about in the other direction? If you create a task in WebCit,
will it display in Outlook? If not, we should probably start generating
new tasks using text/vtodo.
So I'm checking out network replication and it appears that the
global address book can be shared across the network like any other
room. Indeed if I look at the logs for the server it seems that it is
shared on the network regardless of the settings in the room config.
I know there are a few hidden system rooms like the SMTP spool room
but surely not 47 of them.
So what are all these other rooms?
It's a side effect, unfortunately. All rooms which exist in any user's
private namespace (mailboxes, etc.) are counted as being on floor 0.
If I understand licenses correctly, you can convert BSD code to a GPL
license, but not the other way around.
Wow, for the first time ever, I'm having trouble keeping up with what
everyone *else* is doing with the Citadel code. Nice work! :)
Reverted one of the strlen() to IsEmptyStr() changes
because it was causing the indexer to crash.
Well, there it is. I haven't figured out the exact reason why, but the
IsEmptyStr() code is definitely causing that function to crash. It goes
into some weird loop of trying to
here a proposal with a heading which would comprise the Citadel
logo, the site name, the user name and the log off button.
To be honest, I don't really like this ... it seems to take up too much
screen real estate.
If it can be shut off, though, I guess it's an option.
Sure, if you can make it fit, go for it.
ok, glad to hear. i've put in the old function, so it should run
fine too.
I think it is working better now. At least in some simple testing, the
indexer thread no longer hangs, and no longer keeps the server from
shutting down.
This is kind of funny, as chkpw should use all calls citserver will
use, right? or did we forget to put some into chkpw?
I don't know much about Kerberos. Isn't it supposed to do some sort of
three-way handshake that eliminates the need for the user to type a
password? Straight
Hmm, what is it about the old extract_token() that makes it work where the
new one doesn't? Some unintentional side effect that we rely on?
so |foo|bar|bla
gimme number 3
will give you nada?
The tokenizer understands zero-length tokens. So in the example you
gave:
Token 0 is
Token 1 is foo
Token 2 is bar
Token 3 is bla
And if you supply it a string that doesn't contain the
Very nice, thierry. This is going to be much easier to work with now.
Ok guys, we need to fix citserver. It hangs on shutdown (waiting for
threads to complete?) and crashes randomly, and hangs in some other
places. Anyone who recently committed changes, please re-test everything.
Thanks.
As good as the gradient background image in webcit is - its way too
big - 132K.
Surely some compression won't hurt. Switching from a radial to
repeat-y gradient would be better bandwidth wise.
I don't know how to do that ... I just downloaded a canned background
from a
Oops, I spoke too soon. Thierry knows how to do it. Thanks.
The color you changed is the color of of last pixel on the bottom of
the gradient background image, which is not repeat vertically. So if
you have many messages the transition between the gradient image and
the bottom will be really ugly
While I agree with this in concept, the
It means you're opening a unix domain socket.
Current web style
http://www.webdesignfromscratch.com/current-style.cfm
There are some good ideas in here. I'm going to try to use the examples
and guidelines as a reference.
The Citadel build itself probably wouldn't be nearly as involved as making
sure you have all the correct tools installed. No big deal though,
because dothebart's excellent buildpackages script can build a
distribution tarball very quickly on a Linux machine, which can then be
brought over to
So far so good. It built cleanly on my laptop (Ubuntu 7.04), which is
the machine that barfed on it earlier. I normally don't do software
development on this machine, though, so I'll head back over to the usual
dev box and make sure it's still clean there.
To avoid having to
Everyone please go here:
http://tinyurl.com/ysbqfl
FSDaily is a site similar to Digg, but specific to free and open source
software. I've posted a link to our GPLv3 announcement, and would
appreciate it if everyone could vote for this story. It only takes
about a dozen votes to
davew: broken build alert!!
In file included from server_main.c:61:
modules_init.h:1: error: expected identifier or â(â before â-â token
modules_init.h:8: error: stray â#â in program
modules_init.h:9: error: stray â#â in program
modules_init.h:11: error: stray â#â in program
Still doesn't seem to work.
Oooh, the ability to easily create new maintenance threads would be a big
win.
Still broken. I checked out a completely new copy of the tree and attempted
to build it on Ubuntu 7.04 (32-bit Intel).
/mk_module_init.sh
Scanning extension modules for entry points.
This version of echo /bin/echo supports -e -E.
-e Found entry point in file serv_imap.c
-e Found entry point in
For a while, we were experimenting with a way to auto-reconnect clients
through a server restart, but it ended up being really unstable.
Restarting when everyone is idle might be a good idea.
Nope, just the client that ships with the server.
I've just noticed that webcit shows the time as server time when
creating a message and in bbs view when reading a message instead of
local time.
If you have a reliable way of learning the client's time zone, getting it
back to the server, making strftime() aware of it, and doing
And there we go. A righteous hack indeed.
I like the CTDL_MODULE_INIT() framework. Nice and clean. One thing I
really didn't like about the conversion back to a monolithic server was
that we had to name all of the module init functions. This fixes that
very nicely.
If you're inside a module you're already privileged.
Yes, it's true ... some of the modules reference each other's symbols.
There are a couple of situations where I got a bit messy. Sorry :(
1301 - 1400 of 1680 matches
Mail list logo