There is a problem in QW's hull tracing function (inherited from NetQuake's SV_RecursiveHullCheck()) that's causing problems in some places on some maps.

What this particular bug looks like is that sometimes when you're moving along what seems to be a flat wall, you get stopped. The maps themselves are ok; it's the trace algorithm malfunctioning in certain situations.

Here are a few examples: clipbug_e1m1.qwd, clipbug_e1m2.qwd and clipbug_dm4.qwd (that last one is not always reproduceable).

If you know of more such places, please tell me. I have a possible fix and want to see if it works right everywhere.

Oh, and do you think there may be any reason not to fix these bugs?

If I remember correctly LordHavoc already fixed that in his project(s), implemented both client- and server-side fixes, and has a protocol extension for it already etc. etc. You should contact him in my opinion.

Okay, I have committed a the fix to ezQuake. It's essentially identical to LordHavoc's.
Molgrum pointed me to more bugs, on hate and vdm3v3, the fix works for these as well.

Nice nice

presumably you need to add the same code to mvdsv's cmodel.c?

That's right

I will add this to FTESV as well and encourage all server developers to do the same.

gosh i wish i knew what cmodel.c does


a bit too late for me to get involved in qw's programming unfortunately

Will do.