[Gtkradiant] Re: Gtk crashes, off-list email

Timothee Besset gtkradiant@zerowing.idsoftware.com
Sat, 1 Mar 2003 10:42:19 +0100

That should have been an email for the gtkradiant@zerowing.idsoftware.com
list I suppose. 

Build info and stuff is gathered at:

The dependency to cygwin is a bit annoying and overkill. I suppose we
could hack around and take it out, or replace it by a dependency against
python win32 for instance. You can do it and submit a patch if you want,
there's nothing complicated about it really.

There are two sides to vertex edit. We have legacy vertex edit and 'vertex
edit splits faces'. I suppose it's happening during 'vertex edit splits
faces' (crappy code it is). The basic requirement, is that we need a
reproduction recipe. If you get it to reproduce on a win32 build, it
should break with a Crt heap debug assertion. If it's only happening in
release it's trickier, you have to keep a .map file of the release binary
and use that to track down in which function shit happens.. 

Anyway hope I'm giving you enough info to start some serious debugging :)


On Sat, 1 Mar 2003 03:17:59 -0600
"shadowspawn" <shadowspawn@shadowspawn.net> wrote:

> Heyya,
> I wanted to write to you, again early in the morning after some lack of
> sleep; but I'd rather write then forget about what is on my mind. I do allot
> nowadays for modification authors in helping their level designers get their
> stuff working for q3 optimized, sort that unsung 'session artist' that noone
> really speaks of, but there's been a crash that is starting to irk me now
> and I know that it's beyond my capabilities on figuring out where it's
> happening.
> Summation: it's when vertex editing points, and others have seen it too
> which prompts the direct email before putting to the list. I can build a
> debug, but since it's on this hybrid of cygwin on win32 im not sure where to
> send the information to or where the exact cause of the error is occuring.
> This is reported to me on multiple win32's such ,as win9x's and xp and w2k,
> but I have a controled w2k installation that I can offer up to find the
> cause or at least narrow it down.
> I know allot of 3rd party people are starting to use GTK and my interest is
> just to keep people from complaining about it, I ususally just save every
> few changes to a map (snapshots every minute) but some are soon to get irate
> eventually.
> Again sorry if incoherent but I am right now working closely with 3 mods and
> burning the midnight oil so please excuse anything that is mispelled, randy
> from shaderlab can probably explain best what I do with the mod community
> and how I translate to the newest designers what is out there. I had a
> brainstorm and figured I just had to write before forgetting about it,
> because it would help improve the project and I do at the moment have the
> time and resources to offer. Im sure you understand the concept of just
> typing things out and then forgetting about them to free up a part of the
> brain.
> I can build off of the trunk, however I do use the latest 'stable' release
> with production for most edititing (I still use the older one when working
> with entites because of the class change via right click but that's an old
> thing and I am hard headed with that) but I would like to know how to track
> this down with the current cvs, or at least how to offer legitimate
> information to whomever is working on the particulars with whatever area is
> causing the errors. My vaugeness should explain unto itself where I request
> information in order to narrow this down.
> I suppose that I just need a little more integration into the hard-core
> program in order to isolate this and give legit information.
> Wrap-up: This crash happens during vertex point editing in any window
> (camera, x,y,z), and on at least a 16 unit grid, the editor just crashes
> with a dr watson with a generic error about memory alloc, and I'd like to
> contribute more to the project instead of just a 'bug entry', but have no
> idea where to look for detailed information to help isolate the cause or how
> to create a debug build or how to reproduce the steps. I do have a map now
> that is a piece, isolated, but after a dozen or so editing steps I cannot
> accurately reproduce this, but can firmly state that a 6th sense of 'its
> gonna crash soon' does seem to occur right before it does.
> -keith
> aka shadowspawn