map compiling fail

[y2keeth]

tried the official release waited longer this time and it ran but failed the proc and no aas

qconsole.log

[y2keeth]

here are the map files

hanger.zip

[themuffinator]

I checked your log and the compiler. The .proc was written, but its header used a fixed placeholder CRC; that is why the engine immediately called the newly compiled output out of date. acd5d382, included in the 0.13.1 candidate, writes the actual geometry checksum. A local compile of mp/liquid_lab produced .proc/.cm, entered gameplay, and no longer reported the new .proc as out of date.

The missing AAS is a separate, real limitation: openQ4 does not currently contain the Quake 4 AAS compiler implementation. The automatic call was commented out, so compilation silently left navigation unchanged. 0.13.1 reports this explicitly. Changed geometry still needs navigation rebuilt with the Quake 4 SDK tools; reusing retail AAS with a changed map does not give a valid navigation mesh.

I am keeping this issue open for the AAS compiler work. The local runtime check above was a compact map, not a claim that the complete 37-minute Hangar compile has been qualified.

[y2keeth]

ok

[themuffinator]

0.13.1 ships the geometry-checksum fix and an explicit warning that AAS navigation was not rebuilt. The checked dmap output loads into gameplay without the previous false .proc checksum warning.

The remaining work is the missing Quake 4 AAS compiler, so this issue stays open for navigation generation. Map authors still need the Quake 4 SDK tools to rebuild AAS after geometry changes.

[y2keeth]

checking it out now

[y2keeth]

failed qconsole.log
no proc

[themuffinator]

Two separate things were happening here, and both are now on main.

Your .proc was written — it just wasn't where you were looking.

The second log shows the compile succeeding:

----- WriteOutputFile -----
writing maps/game/hangar2.proc
...
 3387 seconds for dmap
...
writing maps/game/hangar2.cm

dmap writes .proc (and now .aas*) under fs_cdpath, which is the directory the engine runs from, while .cm goes under fs_devpath — unregistered by default, so it falls back to fs_savepath. From your search path that means J:\gog\Quake 4\baseoq4\maps\game\hangar2.proc and k:\OpenQ4Data\baseoq4\maps\game\hangar2.cm. Neither is necessarily next to the .map you compiled, which is an easy way to conclude nothing came out.

Every compiled output now prints the absolute path it actually wrote:

writing E:\Repositories\openQ4\builddir\baseoq4\maps\game\hangar2.proc
writing C:\...\baseoq4\maps\game\hangar2.cm
writing E:\Repositories\openQ4\builddir\baseoq4\maps\game\hangar2.aas32

AAS compilation now exists.

dmap rebuilds navigation for every AAS type the game declares (aas32, aas48, aas96, aas128, aas250) straight after .proc/.cm. dmap -noAAS <map> skips that pass, and runAAS <map>, runAASDir <folder> and runReach <map> drive the compiler on its own — handy for your case, since rebuilding navigation does not require repeating the 56-minute geometry compile.

I ran it on the hangar2.map you attached. All five types compile in 16 seconds, and the result lands within about one percent of the retail hangar2.aas32 on every count:

	retail	openQ4
map geometry CRC	4138459670	4138459670
areas	1627	1616
vertices / edges / faces	2875 / 3727 / 9894	2883 / 3712 / 9820
cluster portals	17	17

Loading game/hangar2 with those files reports no AAS warnings at all. Types that no entity in the map uses — aas96, aas128 and aas250 here — get the same empty placeholder file retail writes, so the game recognises them as "this map has no such navigation by design" and loads them silently instead of warning.

One known gap. Raven's tactical feature lump (cover positions, lean and vantage points) is not generated yet. A recompiled map still paths, chases and fights normally; the AI just will not pick cover positions out of the navigation mesh in that map. Retail maps keep their shipped features until you recompile them.

Details and the full parity evidence are in docs/dev/aas-compiler.md.

[themuffinator]

Closing: AAS navigation compilation landed in cad0aa3, and the compiled-output paths are now reported explicitly. Please reopen if hangar2 still misbehaves on your side.

[y2keeth]

still no proc found the cm's but not the proc