WebCit will leave the session open for some time, just in case you come
back. That's just what it does.
You could have it drop idle sessions sooner, if you wanted.
I just checked in some updates to WebCit's RSS stuff. Try it out and see
how you like it!
You mean Firefox doesn't display the RSS feed when you click on it? I
know, go open a bug report with them. GRIN
Er, more specifically, if you just click on the XML button, Firefox SHOULD
give you an Open/Save dialog box, unless you have an RSS reader registered
within Firefox. If that's what you're getting, then nothing's wrong.
GRIN
Hm, now that's odd. Let me try that out.
Why is the RSS reader trying to masquerade as an old version of Firefox?
This is causing a few, er, problems with certain anti-blog spam software.
Hm, I thought I did change it. Maybe I forgot to check it in.
People read README.txt?
OK, I'll hack on the gentoo ebuild sometime this weekend.
I've completely forgotten why that's there, and apparently, failed to
leave a comment. I'll look into it again when I get a few free minutes.
I think I'd rather have the smart terminal from the 80's.
IG, I suggest tzset(3).
Hm, I don't have TZ set.
I suppose you could follow the /etc/localtime symlink and see where it
leads -- assuming it IS a symlink...
Oh yes, it only dates back to 4.3BSD. A decade is relatively new...
I just run all my servers already set to UTC.
Ford, I suggest you rm messages.c and start the whole thing over.
Almost works, anyway. Those URLs don't look right...
I really don't care all that much whether it's libxml2 or expat. Pick one
and I'll get RSS into Citadel sometime the following weekend.
Like this?
r4600 | ajc | 2006-09-22 22:02:03 -0500 (Fri, 22 Sep 2006) | 3 lines
Changed paths:
M /trunk/citadel/housekeeping.c
M /trunk/citadel/msgbase.c
M /trunk/citadel/room_ops.c
M
So which XML parser are we using, again?
Okay, expat it is.
Nov 29 2006 5:04pm from dothebart @uncnsrd
Subject: Re:SVN commit log: revision 4715
You should please do something like that in you citadel commandline
client rewrite too, io-error...
Er, I should do something like what?
And, if you can, please keep the bugfix patches separate from the feature
patches. :)
IG, can you point me to a list of new features in 7.00 (since, say, 6.00)?
This is for generating publicity.
Hm, it SEEMS to work now.
I just submitted the release announcement to slashdot.org. Let's see if
anything happens this time.
Hm, something b0rked in the current svn. citserver won't link:
user_ops.o: In function `CtdlTryPassword':
/.1/home/error/cvs/citadel/user_ops.c:670: undefined reference to
`validpw'
The big things I want to focus on are IPv6 support (for myself!) and
tackling a few of the bits of code that nobody's looked at in a decade or
more.
Eventually I want to put Citadel inside my OpenWRT wireless router. :)
Jan 5 2007 8:23am from IGnatius T Foobar @uncnsrd
As for IPv6 ... I looked into that a while ago and determined that you
can listen on IN6ADDR_ANY (or whatever it's called) and most operating
systems will let you use that to accept IPv6 and IPv4 connections on
the same
Ah, I can't wait for customized CSS.
Thierry's OS X TCP patch (corrected) doesn't seem to have broken anything
on Linux.
And did we ever decide on which XML library we would be using?
expat it is.
On another note, I am having a problem with configure now... it's
complaining about a missing reality distorion field.
IMAP is still a damned abomination.
Now, whichever one of you put in those = signs, please take them back out,
kthx bye.
I don't particularly WANT libcurl; it's too heavy for just fetching the
occasional RSS feed. What else would libcurl be used for?
Whoa, someone touched ctdlphp.
THANK YOU!!
Sweet!
The new format bounce messages have an off-by-one error somewhere.
A message you sent could not be delivered to some or all of its
recipients. The following addresses were undeliverable:
[EMAIL PROTECTED]: User does not exis
I've got dozens of these now, where the
I'm not getting into the cvs vs. svn debate again.
The big thing to know about cookies is that when you send a cookie to the
browser, that you have not yet received the same cookie from the browser
(that happens on the next HTTP request). This affects how you design code
which uses that cookie. If you have just sent someone a cookie with their
Good. That's been driving me absolutely batty for the longest time.
Especially running Citadel on a tiny little VPS.
Now, if only I could find the memory leaks in SpamAssassin...
I've been on vacation for the last week and it's going to take me at least
another week to catch up with all the stuff I normally do. :(
Um, you could just post it here. :)
I read it, and it annoys me. But mainly that's because it reads like too
many cooks were in the kitchen. I can't wait for version 3.1. But I guess
it'll do for now.
Asterisk voicemail?!?
I complained of memory leaks and occasional excessive CPU utilization.
serv_crypto.c really isn't a module in the sense we've been discussing.
This nonce stuff has caused the listsub form to break. It's impossible to
submit the form at all; the nonce never matches.
HTTP: GET /listsub HTTP/1.1
Creating a new session
language found: en_US
HTTP: GET /static/webcit.css HTTP/1.1
Creating a new session
language found:
So why not just use a nice simple mutex instead of the (relatively)
bloated begin_critical_section?
I didn't write begin_critical_section(), I wrote the previous iteration of
lprintf().
I personally log to syslog. So I don't really suffer from these goofball
issues. Maybe I can rip off some code from them or something.
And I have run the server out of memory before. I'd personally like to
have some note of this logged so that I can actually do something about
it,
Oh, I know what's going on with my memory. The VPS is underpowered and
SpamAssassin is a hog.
OK, yes, that can be line buffered, but you might want to switch to
unbuffered when you're reading/writing binary data.
You realize that RSS feed titles can change, right?
Those FeedBurner feeds can be tough on parsers.
Is the Easy Install script still in svn somewhere? I would like to beat on
it a bit to make it do the Right Thing on multilib systems.
Actually I'm about to hack the Makefile(.in) to bits and pieces, since on
my first attempt to make a Citadel RPM, it literally put everything in
/usr . The debian stuff will almost certainly need to be adjusted a bit
once I fix this. G
Oh, it's make install*-new, eh? That explains a lot. Maybe I don't have to
change much after all...
Please don't check in changes which break the build. Thank you.
ssh://uncensored.citadel.org/appl/svn/trunk/libcitadel
Time zones are so horrendous that the usual method of dealing with them is
to store everything in UTC and then to leave displaying the local time to
the client, which presumably knows its own time zone.
Where this has some odd results is with events where the day, rather than
any
Oh, yes, I'm waiting to add Citadel until libcitadel is officially
released or whatever. It's kind of a large change, after all.
I suspect the problem is that res_query isn't reentrant, and so you
(shouldn't|can't) use it in a threaded application.
Oh, I can tell you exactly where it fails. From config.log:
configure:6867: checking for res_query in -lresolv
configure:6902: gcc -o conftest -g -O2 -Wall -Wcast-qual -Wcast-align
-Wstrict-prototypes -D_REENTRANT -pthread conftest.c -lresolv -lrt 5
conftest.c:40:
It's a pretty stock Fedora 8 system. And the new test doesn't work either.
checking for res_query in -lresolv (with resolv.h if present)... no
configure: error: libresolv was not found. Citadel requires the resolver
library.
From config.log:
configure:6931: checking for
I said moved, not removed. I know it needs to be there. I'm just not sure
if the test for res_query should be moved down, or the test for resolv.h
moved up. Or maybe it doesn't matter.
Hey, did somebody forget to check something in? I can't find any file
containing definitions for these:
user_ops.c: In function bsession_startupb:
user_ops.c:546: error: bVIEW_MAILBOXb undeclared (first use in this
function)
user_ops.c:546: error: (Each undeclared identifier is
Stating the obvious, but you did do make install didn't you?
Doh! Sometimes you do have to state the obvious. NOW it works.
CtdlIPCGetMessages() actually returns a NULL-terminated list of the
message IDs in mret. If you want to count them, you certainly can do so.
But the function doesn't return the count for you.
I'll see if I can finally get some RPMs whipped up.
* if we find an '', try to watch out for an ; within the next 6 chars.
if we find a ';' we probably found an entity, and the U RL is finished.
What would an entity be doing in the middle of a URL?
It would be nice if whoever is in charge of the citadel.org DNS pointed
rpm.citadel.org at 64.79.209.100. Then I could make a nice official
repository and everything. :)
Hm, that looks eerily familiar.
Speaking of modules, I liked the module system a lot better when it was
dynamic shared objects. This made it much easier for packaging separate
features separately. The way it is now, I run into dependency hell, having
to install things on people's systems whether they want the features or
Good point. Apache has about 16 different methods of loading shared
objects depending on the OS it's being used on. Too bad we can't just rip
out their code.
Yeah, they weren't my templates.
Aug 25 2008 5:21pm from dothebart @uncnsrd
hm, firefox 3 prints the URL line de-url-encoded though it sends it
encoded...
thats real weird.
I absolutely cannot STAND this. I wish I knew how to turn it off.
OK, now the OpenSSL check is broken.
Nevermind, it was my fault. I didn't have the right path in
/etc/ld.so.conf.
Sep 5 2008 8:16am from IGnatius T Foobar @uncnsrd
Why? What does nginx do that Apache wasn't doing?
Um, among other things:
* Stay up.
* Not leak memory.
* Run fast.
Yeah, I forgot about nginx not taking up much memory. I never see it in
'top' unless my site's really busy. I switched over a month or more ago.
http://linux.ioerror.us/2008/07/migrating-from-apache-to-nginx/
dothebart, no, you just need one PHP fastcgi process. Though all the nginx
people are swearing by php-fpm now.
Cygwin has libintl/gettext/whatever the hell it is, I think. Maybe you
don't have it installed?
Thanks, IG. Now, when can I break backward compatibility on crypto on port
504? There's a potential security issue I need to kill in its TLS
implementation, but Ford will have to recompile his client (so will
everyone else). It doesn't affect SSL/TLS on other ports, as far as I can
tell.
That's the whole change. Aside from renaming the commnad, of course, which
I still think is unnecessary. :)
I'm bck!
OK, so I need to implement a mailing list which only accepts messages from
a specific list of email address, and NOT from any possible subscriber. Is
the best way to do this: to extend the spool control file stuff, stick the
data somewhere else, or something else I can't
Right, the email addresses allowed to post to the mailing list wouldn't have
accounts anyway, and adding authentication to that process would probalby
be impossible.
By writing code to handle this situation.
88 matches
Mail list logo