[Pan-users] Re: Pan "timing out"

[Duncan]

Ron Blizzard <address@hidden> posted
address@hidden, excerpted
below, on  Wed, 03 Jun 2009 16:40:19 -0500:

> Hopefully I've turned HTML off for this response. It looks like it may
> be a toggle -- turn off HTML once and it stays off until you turn it
> back on. But I have no way knowing for certain, as I don't see any of it
> on this side.

You got it this time! =:^)

Well, you got it, but Google's still doing something with it that pan 
doesn't like.  I think it's pan in this case, tho.  It displays fine, but 
when I went to reply, it didn't insert the quote as it normally does!  It 
had the attributions, several blank lines, then my sig, but no actual 
quote!

As it happens I'd seen pan do this once before, and traced it that time 
to an incorrectly specified UTF-8 or whatever character, so I had an idea 
what to look for.  Sure enough, in the "raw" encoding of your post, 
actually, at least once where you had been quoting me (but I checked, it 
wasn't in my original message), there was a quoted-printable escaped 
sequence (= sign in words here so as not to trigger anything weird):

equals C2 equals A0

A bit of googling reveals that that sequence is the (quoted printable 
escaped) encoding for the UTF-8 & nbsp (non-wrapping blank-space) 
character, which means blank space, but keep the words on either side of 
it on the same line, don't wrap at the space to the next line.

Well, pan evidently has problems quoting that extended character 
properly, tho as I said, it displays fine.  But it quotes it properly 
part of the time too, as depending on how I selected text, I could get 
the line including the blank space to come up quoted /part/ of the time, 
but part of the time it wouldn't.  And it had no problem at all with it 
in other paragraphs.  Just that one, and one other.

So what I ended up doing to get the message quoted, was select a portion 
of it that would quote when I hit reply, then select another portion and 
hit reply again, select-copying that text over to the first reply (doing 
that way so the >> indicated quote levels remained correct), then again 
with the next segment, until I had the whole post quoted in one reply.

Now I'm replying to it, deleting as usual the parts I'm /not/ replying to 
as I go.

But as I said, I'm quite sure it's a pan bug as I've seen it before.  
Google is simply triggering it with its (apparently entirely proper) use 
of the UTF-8 non-wrapping blank-space character.

Actually, I suspect it's not even a pan bug, but rather a bug in the gtk 
text widget pan uses.  (I've other reasons to suspect that, that are 
rather too complex to worry about here, except to mention I've 
experimented with attaching yenc encoded text to pan posts, and it 
doesn't work, due to related issues with that text widget.)

Anyway, you and Google have it correct now, AFAICT.  It's pan that's 
bugging now.  Thanks!

>> The (high level) error log (pan calls it the event log)
> 
> I've got that little light bulb -- it never shows any error -- Pan just
> times out. I'm beginning to get the feeling that CentOS hasn't
> implemented Pan very well. Something is not right.

Something is indeed not right, and as you mention below, you can try 
compiling your own pan version.

(FWIW, one thing that's nice about Gentoo is that since it normally 
compiles stuff based on scripts, it's normally much easier to compile 
stuff outside the package manager as well, when necessary, because unlike 
most binary distributions, all the compiling level dependencies tend to 
be already installed.  I remember the hassle I used to have installing 
all the dependencies necessary to compile something on Mandrake...  Of 
course, where that's /not/ needed, just using the pre-compiled binaries 
is easier and faster, particularly on an older or slower computer, but my 
computer is fast enough and has enough resources compiling isn't too bad 
at all, and given the hassles it saves me, Gentoo's definitely the 
correct choice for me.  YMMV of course, and CentOS may well be your 
correct choice, so I'm /not/ pressuring a switch, unless you're looking 
for something different, of course, they we can talk. =:^)

But note that as mentioned above with the weird character issue, it's 
entirely possible it's not pan itself, but GTK or some other pan 
dependency, or something weird in the environment.  For example, with the 
help of various users on this list, we eventually traced a terrible 
performance problem with pan on Ubuntu, to the GNOME assistive 
technologies (AT) applet, of /all/ things!  Who would have thought, 
especially when pan is a GNOME family app?  But that's what it was, and 
disabling it killed the problem with pan.

> I may try the --debug switch. But I'm beginning to wonder if the fight
> is worth it.

Indeed.

> It doesn't appear that there are any errors in Pan -- it just loses the
> connection. Once that happens, I get an error 441 -- Newsgroup Can't be
> found. The headers remain and I can click on them, but I just get a
> blank page. The only "fix" is to exit Pan and restart it.

I still say that sounds like too short a TCP idle-timeout on your 
router.  But again, if pan's the only thing running into it, honestly, it 
may be easier using something else instead of pan, than fiddling with the 
router or whatever else.

> Unfortunately the "L" doesn't seem to help.

That sucks.  However, after I posted that, I had a slightly different 
issue here with my sucky ISP's server and needed to cut the (still 
working) connections and restart them.  I /thought/ that once I hit the 
offline button, pan would terminate the connections.  But I thought 
wrong.  Turns out that function doesn't work quite as I thought.  It only 
tells pan not to fetch anything.  It doesn't (apparently) tell pan to 
terminate existing TCP connections, so it didn't help my problem and 
wouldn't help yours either, unfortunately.

I was ultimately able to get the connections killed so pan would start 
new ones, not using pan, but by shutting down and restarting the 
machine's network interface.  Since I tested that and it works here, its 
very likely that it would work for you too.

However, not only is that honestly rather much to be asking, particularly 
since it means killing any unrelated connections (here, very often it 
would mean killing a shoutcast/icecast incoming music stream, who knows 
what there), but even if you're willing to do it, I'd have to hope you 
know CentOS well enough to do so as I haven't the foggiest how it's 
handled there, whether it's a manual ifconfig down, ifconfig up, or 
whether you restart a service, or what.

> It's becoming clearer to me that it's probably not a Pan problem. Maybe
> I'll look into downloading a newer version of Pan and compiling it.

It's worth trying, especially since it may mean omitting RedHat/CentOS 
patches that may be proving troublesome, but as I said above, even if 
it's pan, it may really be GTK or some other component pan uses, not pan 
at all.  So you might end up having to rebuild that too... all for an 
iffy at best chance of fixing it, since I still strongly suspect it's the 
router's TCP idle timeouts.
 
> Reflashing my Linksys router will probably be a "last resort" kind of
> thing. I would rather limp along with Thunderbird than take a chance of
> turning my router into a brick.

Entirely understood.  I simply presented it as one option, knowing it's 
one some aren't going to want to take.

>> BTW, [t]here's malware going around that tries to bot those routers
> 
> Thanks for the tip. I'll look into this.

FWIW, I came across a couple of the links again (reading the notes on 
openwrt.org, the 0th link below) and thus have them handy:

0. http://openwrt.org/  (down the page a bit, or search "Botnet")
1. http://it.slashdot.org/article.pl?sid=09/03/23/2257252
2. http://dronebl.org/blog/8

I believe I saw it first on The Register, but don't have a link for that 
one.

> I'll try telnetting. That was the way I used to have to program my old
> router. (Kind of forgot about it.)

Telnetting is a very useful troubleshooting skill to have.  Among other 
things, you can telnet a web server (normally) on port 80, or a news 
server on port 119, or mail servers (pop3 and smtp at least, not sure 
about imap), just to see what the connection banner is, even if you don't 
know the necessary commands to actually do anything.  (quit or QUIT 
usually works to terminate the connection, and often the help or HELP 
command still prints out the other available commands, tho manual access 
is rare enough these days the help command is beginning to disappear as 
well.)  Just knowing and using that has been /very/ useful for me at 
times, particularly when the error the GUI app is reporting is entirely 
unhelpful, as the bare protocol error reported by telnet if it gets one, 
usually at least makes sense, and just knowing for sure whether telnet 
can make the connection or not, goes a long way on its own.

> Thanks. You've given me a few more directions to go. I'll try to respond
> more quickly to any new messages.

Don't worry too much about it.  It's your problem, not mine, so your 
timeline working on a fix, not mine.  Of course, if you wait six months 
or something, I may well have forgotten and we may need to rehash some 
stuff, but a month or six weeks, not an issue at all.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman