RPCS3
[Regression] Resistance FOM: Crash on Demuxer since build v0.0.9-10115 (#7952)
Quick summary
Follow up of #16563 to narrow it down.
The crash when the reproduction of the intro video is started (after waiting on menu without pressing on start button for about 60 seconds) is a regression appearing in Resistance FOM since v0.0.9-10115 (#7952). SPU Decoder set to Interpreter (fast) was needed to start the game.
As reported in #16563, the issue seems due to an attempt to allocate memory no more available on the lv2 container. The check for memory availability failed (so no memory was allocated) and that will cause the crash below:
F {PPU[0x1000046] Thread (Demuxer) [0x005ec2cc]} VM: Access violation writing location 0x0 (unmapped memory)
Removing in the code the check on residual memory availability in lv2 container fixed the issue but of course only developers can investigate the reason of such memory need just to reproduce a video (e.g. memory attached to lv2 container previously not properly released? etc.).
Details
No response
Build with regression
Working build: v0.0.9-10114 Regressing build: v0.0.9-10115 (#7952)
Attach two log files
Attach capture files for visual issues
No response
System configuration
No response
Other details
No response
UPDATE:
I found the build responsible for the regression. It is build v0.0.9-10115 (#7952), I needed to use SPU Interpreter to start the game. That PR is one of the first PR updating global memory stats.
Updated ticket description and new log files.
UPDATE:
As a workaround (hopefully temporary) it seems that enabling Debug Console Mode (adding extra 0xC000000 bytes to the container) on Advanced tab fixes the issue.
This is still an issue on latest. Debug Console does work as mentioned.
Does it work on libvdec.sprx?
I also tested this in the past. It freezes with a black screen when going on main menu (so when loading the fist video; not the one after 1 minute of inactivity)