Re: w32: buglike behavior ver. 22.0.50.1 on msw98

[Eli Zaretskii]

> From: <address@hidden>
> Date: Fri, 14 Apr 2006 20:05:43 -0500
> 
> Version is: GNU Emacs 22.0.50.1 (i386-mingw-windows98.2222) of 2006-01-29
> on NEUTRINO
> 
> Thanks,
> 
> Ed
> 
> p.s. I do have the gdb.exe for windows and a lot of cygwin tools but I
> don't know how to use them.

Can you run Emacs from GDB, do whatever you do to trigger this bug,
and then see whether GDB kicks in at some point?  If it does, we may
be lucky enough to figure out what goes wrong.

To start Emacs from GDB, run GDB from the src directory of the Emacs
tree, with the command "gdb ../bin/emacs.exe", then inside GDB type
"run -Q".  (This assumes that emacs.exe is in ../bin relative to the
sources; if not, modify the GDB invocation command line accordingly.)
If you cannot reproduce the bug with -Q, try to throw together the
smallest .emacs that reproduces the bug, and invoke Emacs with just
"run" inside GDB.

Btw, please don't take this discussion off emacs-pretest-bug, as there
are other Windows developers here who could help with ideas.  I'm
reproducing your private message to me below for them to see what you
wrote.

> > Can you try to determine why it locks up? is it because the resulting
> > buffer is too large for the available memory?
> 
> I don't think I'm enough of a power user to do that.  I did load 2 buffers
> of random bytes (each 32meg+) into 22.5 and it didn't choke on them. In
> fact it loaded them much faster than 21.3 version. After typing C-h a
> command the *Apropos* buffer fills up and, for a while, I can move point
> around and see thousands of CJK ideographs, etc. but later it freezes. If
> I keep trying to click on menu bar, move cursor with mouse slider or M-x-<
> etc. it will blue screen (last time with this, maybe not much help):
> 
> Exception 0E at 0028:C000408D in VxD VMM(01) + 0000308D  THIS WAS CALLED
> FROM 0028:00000000 IN VxD (this on blue screen). Press to continue...
> Normal looking Emacs frame but with black background. It doesn't respond
> to mouse or keyboard. Usually the process can be killed but sometimes I
> find that Windows 98 has been corrupted afterwards and I have to
> Ctl-Alt-Del.
> 
> >
> > > Shouldn't this line be truncated after the ascii range or just list
> > > the ranges of self-inserts rather than each character?
> >
> > The problem is to know where it becomes dangerously large.
> 
> Maybe it's displaying all the glyphs for all the input methods and all the
> different renderings, etc. Many empty rectangles show also not just the
> character ranges arialuni.ttf happens to cover. I would think that the
> buffer should be well under 64 megs unless it's also generating all of the
> permutations for non pre-composed glyphs. Would it help to diagnose this
> if leim were somehow disabled?
> 
> Version is: GNU Emacs 22.0.50.1 (i386-mingw-windows98.2222) of 2006-01-29
> on NEUTRINO
> 
> Thanks,
> 
> Ed
> 
> p.s. I do have the gdb.exe for windows and a lot of cygwin tools but I
> don't know how to use them. I just play around with some filters (sed,
> grep, etc.) from the dos shell.
> 
>