Does anybody know whether Ironwail supports checksum-based loading of ent files? IIRC, QSS had that feature and it was quite useful. 

@metlslime :

Fitzquake had bumped MAXALIASVERTS and MAXALIASTRIS practically to twice
their original values (literally in MAXALIASTRIS case) but corresponding
arrays in gl_mesh.c remained the same -- namely the globals commands[],
stripverts[], striptris[], used[] and vertexorder[], and also the locals
besttris[] and bestverts[] in BuildTris().

Is that an oversight?

With current array sizes we get a segmentation fault in mj4m4 and mj4m5
maps of the mjolnir mod but not if we double sizes of those arrays (the
commands[] array in particular.) 

I guess it was an oversight. I did a search and found this message:

https://www.celephais.net/board/view_thread.php?id=1&start=3894&end=3894#3894

Which indicates my motivation to bump MAXALIASVERTS to match software quake's max (which was higher than glquake.) 

Version 0.96.0 of QuakeSpasm is released.

Downloads:
https://sourceforge.net/projects/quakespasm/files/
http://quakespasm.sourceforge.net/download.htm

Changes since the previous version:

- Adjustments to joystick defaults and behaviour. (See the
"Controller support / Cvars" section in the README.)
- Fix for a mouse grabbing issue on macOS.
- Add missing support for -ip command line option on unix / macOS.
- Backport angled sprites code from FTEQW (spriteframetype: 2, must
have 8 frames per group.)
- Bump the progs execution runaway loop limit to 16M.
- Add signon buffer fragmentation to fix SZ_GetSpace errors on
certain maps when using protocol 999.
- Increased MAX_DATAGRAM to 64000 to overcome packet overflows.
- Bump the MAXALIASFRAMES limit to 1024.
- Fix arrays to have proper sizes in gl mesh code, reflecting
MAXALIASTRIS and MAXALIASVERTS correctly. (fixes SIGSEGV in
mj4m4/mj4m5.)
- Increased default value of gl_farclip to 65536.
- Increase chase cam target trace distance to allow for chasecam to
function in large open maps (useful e.g. for the func_vehicle mod.)
- Raised default maximum number of particles to 16384 (can be set
up to 32768 with -particles on the command line.)
- Fix on-screen keyboard showing up suddenly on Steam Deck when
starting a map.
- Fix viewmodel interpolation with >10Hz animations.
- Fix a memory leak when exiting game in SDL2 builds.
- Fix a possible crash when handling clipboard data on macOS.
- Fix a potential crash after loading of saved game.
- Fix possible out-of-bound reads when handling progs type sizes.
- Fix Dutch angle VP_PARALLEL_UPRIGHT sprites.
- Fix an issue with lights blending by using 10 bit color depth for
lightmaps. (For GLSL mode only.) Disable with "-nopackedpixels",
if necessary.
- No relative motions when the window is not focused.
- Status bar and intermission screen tweaks.
- Properly display monster counts > 3 digits.
- Fix console animation with scr_conspeed <= 0.
- Fix mouselook not working upon game quickload during a demo play.
- Haiku operating system support.
- Fix console when server is run noninteractively with stdin
redirected from /dev/null.
- Updated Visual Studio project files.
- Minor code cleanups. Updated third party code, e.g. SDL, music
codecs, etc.
- Thanks to Andrei Drexler, Alexey Lysiuk, Andrey Budko, Boris I.
Bendovsky, Chris Cowan, Maciej Olędzki, Simon McVittie, OscarL,
Eric Wasylishen, Jaycie Ewald and Spike for patches. 

While the app works great in emulation mode with Rosetta 2, I'm wondering if a native Silicon version will come one day. 

M1 requires Apple 'notorization' and I don't have an Apple certificate to sign my builds (the unsigned M1 versions will refuse to run..) The Xcode project file included in the source supports targeting M1, so you can build yourself (or keep using the x64 version.) 

While I don't know much about M1 compilations, I believe there's a way to build native apps for Mac Silicon without the pesky Apple notarization. I may be wrong, though.

Anyway, multi-thanks a lot for keeping Quake alive on Macs! 

believe there's a way to build native apps for Mac Silicon without the pesky Apple notarization.

If there is, I'd like to know.

Anyway, multi-thanks a lot for keeping Quake alive on Macs!

You're most welcome. 

Version 0.96.1 of QuakeSpasm is released.

Downloads:
https://sourceforge.net/projects/quakespasm/files/
http://quakespasm.sourceforge.net/download.htm

Changes since the previous version:

- Fix demo recording as client-only after connection to server (was
broken by signon changes in 0.96.0. Thanks to Jozsef Szalontai for
issue report.)

- Fix potential buffer overflow in COM_Parse(), e.g. with maps with
oversized 'wad' fields. (Thanks to Andrei Drexler.)

- Minor code cleanups. 

since v.0.96 is able to complete Mjolnir for now. May be it is time to update IronWail wich unable to run this mod normally? (map mj4m4 IW won`t load at any way for me). BTW IW my common Quake engine for now. I make no claims, I appreciate work of mr. Andrei Drexler. Thank You. 

(2024-march-01) Every step up the stairs is the pain for eyes, even 8 units high? 

@metlslime, @mh :

I was reported a stack buffer overrun in Mod_PolyForUnlitSurface
https://github.com/sezero/quakespasm/issues/100
Fixed easily by removing the hardcoded local verts and inlining
'convert edges back to a normal polygon' code into poly creation
as shown in the patch in there.

However, a similar local copy to a hardcoded local verts[64] is
also in GL_SubdivideSurface, and the same trick can not apply to
it. Its subroutine SubdividePolygon() checks for numverts > 60
and errors out if it is so, but GL_SubdivideSurface itself does
not check before copying to the stack buffer verts[64]: What is
a good solution here? Easy way out by erroring out if fa->polys->
numverts > 64 ?? 

I don't see an easy way to remove the arrays, since it's recursively creating new arrays as it goes. You could throw an error in GL_SubdivideSurface like you suggested. or do a warning and drop all of the remaining verts (leaving an incomplete polygon.) 

OK, will throw an error. Thank you. 

if surfaces are always convex polygons, GL_SubdivideSurface can pre-divide it into a number of “triangle fan” parts. that could be done inline by messing with the vertex array in place and keeping track and fixing it up as you go. then youre only sending safe data into the recursive function. 

If you can send me a patch (sezeroz _at_ gmail _dot_com) I can
gladly test and include it in the next release. If not, well, we
have never hit a stack overrun in GL_SubdivideSurface so far and
this was just a safeguard anyway. Thanks. 

Version 0.96.2 of QuakeSpasm is released.

Downloads:
https://sourceforge.net/projects/quakespasm/files/
http://quakespasm.sourceforge.net/download.htm

Changes since the previous version:

- Fix stack buffer overrun Mod_PolyForUnlitSurface: fixes crash when
loading lim_daviddg.bsp from Liminal Spaces Jam.
- Fix potential buffer overflow in progs global string functions.
- Fix potential buffer overflow in Mod_LoadAliasFrame()
- Optimize BoxOnPlaneSide()
- Reserve enough space in PF_makestatic() for worst case scenarios.
- Improve handling of huge entities in SV_FindTouchedLeafs()
- Warn about bad maps with only 1 or 2 edges per face (e.g.: mj4m?,
alk_dancing, ej3_bizz, rotj_entsoy...)
- Apply FOV gun scaling for r_showtris
- Sliders with visible values in options menu.
- game command now refuses setting a non-existing game directory
- Fix build in C23 mode.
- Minor code cleanups. Updated third party code, e.g. SDL, music
codecs, etc.
- Thanks to Andrei Drexler, Alexey Lysiuk, Diordany van Hemert, and
Jaime Moreira for patches. 

Version 0.96.3 of QuakeSpasm is released.

Downloads:
https://sourceforge.net/projects/quakespasm/files/
http://quakespasm.sourceforge.net/download.htm

Changes since the previous version:

- Fix potential infinite loop when playing sounds with loop start >= end
(e.g. misc/forcefield.wav from Madfox's kaptlog.zip)
- Fix possible overflow when resampling very big sounds (e.g. gram1.wav
from the 'A day like no other' mod.)
- Change Sky_ClipPoly to allocate on the heap in if MAX_CLIP_VERTS limit
is reached (e.g. with lim_daviddg from Liminal Spaces Jam)
- Minor code cleanups. Updates to third party code.
- Thanks to Andrei Drexler, Vincent Sonnier, Alexey Lysiuk, and Jaycie
Ewald for patches. 

any particular reason the

-"elevator sinking bug", (plus airquake tank turret updating bug)
-the "stuck on convex angle bug",

still not fixed after all these years?

...at least you guys fixed the mouse glitch bug 

Some bugs are content bugs in the maps, not engine bugs. 

i don't recall these bugs happening in winquake

particularly the "elevator sinking bug" (and airquake tank turret updating bug) 

the "elevator sinking bug" doesn't happen in it... 

Stuck on convex angle is 100% reproducible in ID Quake and happens in e1m1; it was fixed in the Kex reissue by altering the map geometry. 

I'm looking at this page but couldn't find such an option, does it exist? Thanks!

https://manpages.ubuntu.com/manpages/bionic/man1/quakespasm.1.html 

just adding “+map mapmame” should do it