The not so useful backtrace I have so far: Thread 1 "WPEWebProcess" received signal SIGSEGV, Segmentation fault. 0x7092acf8 in ?? () (gdb) bt #0 0x7092acf8 in ?? () #1 0x7674f9c0 in operationStrCat3 () from target:/usr/lib/libWPEWebKit-1.0.so.3 #2 0x00000018 in ?? () Backtrace stopped: previous frame identical to this frame (corrupt stack?)
The crashing benchmark seems to be WSJ. Disabling the DFG JIT works around the issue but renders the whole thing unusable because too slow.
I'm investigating it.
It looks like a bug on tail calls corrupting stack on some cases. I'm trying to get a simple case to reproduce the bug, but I was able to run "WSL" on JSC command line setting "--useTailCall=false".
Any update in this problem, we are checking in the rpi3 and the test crashes very early.
WSL issue should be fixed after https://trac.webkit.org/changeset/258143 landed. I'm able to run it into JetStream 2 CLI. However, due to memory constraints, I'm unable to run the entire JetStream2 without OOM in my rpi device.
Just checked, the WSL crash is fixed for me in the rpi device too.