TNT not working reliably in Emacs

TNT not working reliably in Emacs

[Scott Evans]

Hi all,

I'm one of the maintainers of TNT, an AOL Instant Messenger client that
runs in emacs.  (http://tnt.sourceforge.net -- right now only the CVS
version is usable.)  I've been an XEmacs user for a few years, and I'm in
the process of switching over to Emacs.

I -- and some other TNT users -- have noticed that TNT is not reliable in
Emacs.  For instance, logging in to AIM succeeds on, I'd say, 10-20% of my
attempts.  On XEmacs that number is closer to 100%.

I've spent some time trying to chase this down, and so far I can't find
anything to point at.  Both successful and unsuccessful Emacs logins are
sending the same login info (using process-send-string)... but sometimes
a login response comes back (via a process filter), and most times one
doesn't.  I've not yet tried a proxy to see if the server is responding
in the unsuccessful cases.

I *have* tried this using NTEmacs 21.3, NTEmacs 22.0.50.1, and 21.4.1 on
debian; all three behave similarly.

Are there any known, perhaps longstanding, quirks with process-send-string
or process filters?  Would anybody be interested in checking out the TNT
code to see if it does something that's just plain wrong?  I'd be happy to
point folks to the pieces of code in question.

Thanks-

gse


--
 scott evans :: www.antisleep.com

Re: TNT not working reliably in Emacs

[Richard M. Stallman]

      I've not yet tried a proxy to see if the server is responding
    in the unsuccessful cases.

It would be very useful to find a way of determining exactly
what the server sends back.  Then we would know for certain
whether the data is lost inside Emacs.

If it is, we could add some debugging code to log everything inside
the relevant asynchronous Emacs code, and that way we could fix the
bug.

Re: TNT not working reliably in Emacs

[Scott Evans]

On Sat, 31 Dec 2005, Richard M. Stallman wrote:
>     I've not yet tried a proxy to see if the server is responding
>     in the unsuccessful cases.
>
> It would be very useful to find a way of determining exactly what the
> server sends back.  Then we would know for certain whether the data is
> lost inside Emacs.
>
> If it is, we could add some debugging code to log everything inside the
> relevant asynchronous Emacs code, and that way we could fix the bug.

That would certainly be good.  I'll look into proxying over the next week
or two and see what I can find.



gse

TNT not working reliably in Emacs

[Scott Evans]

Hi all,

I'm one of the maintainers of TNT, an AOL Instant Messenger client that
runs in emacs.  (http://tnt.sourceforge.net -- right now only the CVS
version is usable.)  I've been an XEmacs user for a few years, and I'm in
the process of switching over to Emacs.

I -- and some other TNT users -- have noticed that TNT is not reliable in
Emacs.  For instance, logging in to AIM succeeds on, I'd say, 10-20% of my
attempts.  On XEmacs that number is closer to 100%.

I've spent some time trying to chase this down, and so far I can't find
anything to point at.  Both successful and unsuccessful Emacs logins are
sending the same login info (using process-send-string)... but sometimes
a login response comes back (via a process filter), and most times one
doesn't.  I've not yet tried a proxy to see if the server is responding
in the unsuccessful cases.

I *have* tried this using NTEmacs 21.3, NTEmacs 22.0.50.1, and 21.4.1 on
debian; all three behave similarly.

Are there any known, perhaps longstanding, quirks with process-send-string
or process filters?  Would anybody be interested in checking out the TNT
code to see if it does something that's just plain wrong?  I'd be happy to
point folks to the pieces of code in question.

Thanks-

gse


--
 scott evans :: www.antisleep.com

Re: TNT not working reliably in Emacs

[Scott Evans]

> > Are there any known, perhaps longstanding, quirks with
> process-send-string
> > or process filters?  Would anybody be interested in checking out the TNT
> > code to see if it does something that's just plain wrong?  I'd be happy to
> > point folks to the pieces of code in question.
>
> Show us the code.

Okay.  Download a copy of my latest here:
  http://www.antisleep.com/tnt.tgz

There's a lot of debug in these files, from me trying to trace through and
see what's happening.  All of the debug output is placed in a buffer
called *gse-debug*.  (a clean, mostly-sans-debug version can be obtained
from sourceforge CVS.)

You'll need an AIM account (preferably with a few buddies) to really test
this stuff, since the problem occurs while logging in.

As far as I can tell, the login string that is sent to tocstr-send-flap1()
is consistent in success/failure cases.  That's why I wonder if
process-send-string(), or Emacs' underlying filter code, is not working
right.  But - who knows, I may be chasing the wrong thing entirely.

Any help is greatly appreciated!


gse