Christian Despres

Since nothing is ever deleted from the JITServer AOT caches, we could even have a static `_totalRecordBytes` in the `AOTCacheRecord` class that keeps track of this data, perhaps with its...

That makes sense, as a first pass at the problem. Is setting `_aotCacheStore` to false the only thing that needs to be done? If so, then we'd only have to...

@mpirvu. I'll try to reproduce this locally. From the recent commits I noticed 1a92d685270a85d9ba369ae872dcbeebea0620d3 that touches the JITServer network communication, but I haven't looked at it in detail.

I am able to reproduce this locally.

After bisection it appears that the bad commit is in fact d505324ff4b258f3ddfd49e23a67b3c59917d713.

Setting ` -XX:-JITServerLocalSyncCompiles` does seem to solve the issue.

I can generate a client verbose log by setting `EXTRA_OPTIONS`, but the segfault does not occur when I do.

It appears to run in client mode still: ``` #JITServer: JITServer version: 1.38.0 #JITServer: JITServer Client Mode. Server address: localhost port: 38400. Connection Timeout 30000ms #JITServer: Identifier for current client...

If it's time dependent, would you like me to try to get a log on a perf machine or something with a bit more power?

I think that this may still be happening, but with different tests now. I see various failures on ppc64le only, though perhaps the x86_64 tests haven't finished yet. (EDIT: it's...