[1.17.x] Crash when trying to run forge forge client

[Yanny7]

Clean build of forge (#94389d0e) is crashing when trying to run 

forge forge client

 

[14:09:38] [Render thread/INFO]: Setting user: Dev
[14:09:38] [Render thread/INFO]: Backend library: LWJGL version 3.2.2 SNAPSHOT
#
# A fatal error has been detected by the Java Runtime Environment:
#
#  SIGSEGV (0xb) at pc=0x00007fcbd89b1112, pid=47207, tid=47208
#
# JRE version: OpenJDK Runtime Environment (16.0.2+7) (build 16.0.2+7)
# Java VM: OpenJDK 64-Bit Server VM (16.0.2+7, mixed mode, tiered, compressed oops, compressed class ptrs, g1 gc, linux-amd64)
# Problematic frame:
# C  [libglfw.so+0x17112]
#

- tried on Arch linux

- Note that I havent problem running it on 1.15.x and 1.16x

- debug.log failed to upload

latest.log hs_err_pid47207.log

[Xograywolfox]

This is happening to me too! Mine is run on win10! Everything was working fine until the new update released. It crashes out; code 0

[Yanny7]

Another info: 

clean clean client

just run without problem

[Yanny7]

I didn't solved that, the reason why it crash is for me unclear... but sometimes it fail in different place, while setting FPS (I debugged that, but it goes thru that methot multiple times without problem, then it failed) .... strange

[malte0811]

I guess I'm a bit late to the party, but I spent essentially the week after the initial Forge release debugging this. At the time it seemed to be a "me" issue, so I didn't really put any "force" behind anything related to it.

1. It only seems to happen when there are "many" files (outside of jars/zips) that need to be scanned, for some definition of many. Try launching e.g. the 1.17 branch of AE2 from an MDK; I've had the same issue with the (private) 1.17 branch of IE.
2. In Forge dev, put a breakpoint here and check the number of stream elements on the last invocation (third for me). Kill MC, limit the stream to that number of elements in the last invocation (just using a global counter) => no crash (at least for me). Limit the stream to one more element => crash again.
3. I have a "highly heuristic" fix: Build this PR to securejarhandler and replace Forge's SJH dependency with it. At least for me this reliably fixes the crash, in Forge dev as well as in IE dev. I have no clue why it works, whether it's a fix or a workaround and if it will work for anyone else or just me (so I'd be interested in data on that last point).