Gunnar Evermann <ge204(a)eng.cam.ac.uk> writes:
Jan, sorry for interrupting the work on your thesis.
OK, for one second :-)
for 21.1.7 the following patch was applied:
1999-08-28 Jan Vroonhof <vroonhof(a)math.ethz.ch>
* menubar-x.c (make_dummy_xbutton_event): Don't add
offset twice if HAVE_WMCOMMAND is defined.
http://www.xemacs.org/list-archives/xemacs-patches/199908/msg00080.html
I don't pretend to understand this, but HAVE_WMCOMMAND is definitely a
21.2'ism.
This soooo stupid.. The thing is called HAVE_SESSION in 21.1.x. I
could have sworn that change was between 20.4 and 21 nut it was not..
I can fix the problem by removing the whole
#ifdef HAVE_WMCOMMAND block from 21.1.8 (of course we could then
eliminate shel[xy] altogether).
I don't pretend to grok it fully either. Here is my "guesswork" on
which the patch is based.
The problem here is that the HAVE_SESSION/HAVE_WMCOMMAND works by
making the shell window invisible and keeping a single one. It then
creates a new window to wrap each frame. Each of these gets a copy of
the arguments of the shell window.
Thus in this case we must NOT add the offset from the shell and the
frame together because then we count the initial offset twice. However
if HAVE_SESSION/HAVE_WMCOMMAND is not set then the offset is only
contained on the shell-window and we must copy it from there.
suggestions?
Change the HAVE_WMCOMMAND to HAVE_SESSION. It works for me...
BTW: I still can't build 21.1.9pre on Irix5.3 with gcc... :-(
$@$@#! The dumping patch was intentionally changed to ONLY affect Irix
6.x. This under the assumption that we wouldn't want to break Irix5
which worked with Alexandre Oliva's patch. You are saying that isn't
the case :-(?
Jan
P.S. Did you see that one of the latest incoming bugs contains a
reproduction recipe for a D&D crash (maybe similar to the one that has
been bugging us all the time).