Closed ghost closed 3 years ago
yeah, dmap hasn't been tested much yet, as far as I remember.. I wonder if this also happens in 32bit builds of dhewm3 (might be a 64bit issue)?
I can compile a 32-bit build later today and try with that.
so have you tried 32bit builds yet?
Hello,
Sorry about the lateness! Been busy with things. Unfortunately no, I haven't had the chance to try yet. I've got a laptop running 32-bit Linux which I meant to test it with, but it's at my parent's place which I haven't managed to visit yet in the meantime.
any update on this?
I can't reproduce the problem with latest dhewm3 (from git) on XUbuntu 18.04 x86_64 (GCC 7.5 and clang 6.0).
Also works with XUbuntu 20.04 (GCC 9.3 and 10.2) so I guess this issue has been fixed accidentally.
Turns out I can reproduce it after all, but only if ONATIVE
is enabled in CMake - this would've been relevant information.
https://github.com/RobertBeckebans/RBDOOM-3-BFG/issues/436 discusses the issue, which is caused by a GCC compiler bug (at least I think it's a bug).
GCC bugreport for this: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100839
is -ffp-contract=off
the solution?
I don't know, maybe. TBH all this doesn't make sense to me and I hope that the guy who answered there was mistaken and GCC doesn't really believe that breaking the language standard in normal optimization modes is acceptable.
I'll wait a bit, hoping that someone else chimes in in the GCC bugreport to clear this up. For now my workaround should work for all GCC versions including their current development version.
I've started making maps for Doom 3, using DarkRadiant. I compile them using the 'dmap' command from Doom 3.
The problem is some geometry is rendered wrongly in-game. Here's a thread I created on the idTech4 forum, that demonstrates the problem.
The interesting thing is that these glitches aren't there when the map is compiled in stock Doom 3.
Please take a look at this testmap.