[Xcb] XCB enabled libX11 and tray

[Jamey Sharp]

On Sat, Jun 17, 2006 at 10:38:21AM +0200, morgoth6 at box43.pl wrote:
> Jamey Sharp wrote:
> > Can you run kadu (or kicker) under GDB, set a breakpoint in
> > _XIOError, and get a backtrace?
> 
> Hmmm, I guess I need to recompile libX11, XCB && Kadu with debug
> symbols ? This machine is a bit slow but I can try.

Kadu's debug symbols won't help me, so you don't need to bother. I think
GDB ought to give you good enough results without any debugging symbols
really; but it would be better to have them for libX11 especially. Try
first without recompiling so you can get the results to me faster. :-)

> Hmmm, maybe this can be a problem because I doesn't use regular x86
> machine. I have Pegasos II computer with 1GHz PPC CPU, 1G of Ram. Can
> this be endianese related for eample ?

I don't think so. I'm pretty confident that this code is endianness
clean, though I admit it's a lot less tested on non-x86. :-) I doubt any
other architecture-specific thing is the cause either. But it's a good
piece of information for me to keep in mind, just in case.

I should write down my hypothesis-so-far. I'm drawing inferences in two
directions and hoping they'll meet in the middle.

The immediate symptom of the bug is that apparently _XIOError is being
called, though as I don't think that's the standard error message,
possibly the KDE libraries register a custom error handler. Anyway this
can occur when waiting for a reply, in _XReply, or an event; or it could
occur when trying to send a request.

This bug occurs sometime after the Kadu "window" (== icon) in the kicker
tray receives an EnterNotify event. One of the things Kadu presumably
will do in response is pop up a tooltip; possibly this bug occurs
somewhere in that handling.

Both parts are vague: I don't have enough information yet. The Ethereal
and GDB traces I asked for should be particularly informative.

By the way, are you using XIM (X Input Methods) to input characters that
your keyboard doesn't have? Checking what calls _XIOError reminded me
that the XIM code paths in my libX11 are probably completely untested.

> > This is certainly odd. I don't know if I'll be able to find the bug
> > before I leave, but I can at least ask some questions. :-)
> 
> I hope so :) How much time we have ?

A few days; I'm not sure exactly. Then I'll be gone for a week or so,
back for maybe a few days, and gone for maybe another week. There isn't
exactly a "schedule" for this trip.

I'd like to keep this discussion on-list, by the way. Others may have
insight or the same problem.

--Jamey
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://lists.freedesktop.org/archives/xcb/attachments/20060617/b743c7b9/attachment.pgp