Deduplicating might hang resulting in a broken workspace

It seems I am getting more issues recently with deduplicating hanging in my workspace.

INFO  tracing is disabled for protocol BSP, to enable tracing of incoming and outgoing JSON messages create an empty file at /home/tgodzik/.cache/metals/bsp.trace.json
INFO  Connected to Build server v1.4.0-RC1
INFO  time: indexed workspace in 0.61s
INFO  Deduplicating compilation of input-test from bsp client 'Metals 0.7.6+707-c2394b32-SNAPSHOT' (since 4h 57m 57.571s)
INFO  Deduplicating compilation of interfaces-test from bsp client 'Metals 0.7.6+707-c2394b32-SNAPSHOT' (since 4h 57m 57.575s)
INFO  Deduplicating compilation of metals from bsp client 'Metals 0.7.6+707-c2394b32-SNAPSHOT' (since 4h 57m 57.601s)
INFO  compiling mtest (12 scala sources)
INFO  time: compiled mtest in 0.16s

@jvican Any idea what is going on or where to look?

Killing the server usually works.

[jvican]

The safe lazy null issue reported by @tpolecat is now fixed by https://github.com/scalacenter/bloop/pull/1265. I need to add another fix to Bloop so that when there’s an exception, the compilation that is tracked and deduplicated fails altogether. It should be working like that, I don’t know why it doesn’t.

@matthughes Regarding your issue, it seems different from the one reported here. If you see, you have at one point 6 different clients connected to bloop and asking for that compilation (that’s what the [D] Increasing counter for /Users/mhughe008/development/vise/project/.bloop/commonJVM/bloop-internal-classes/classes-bloop-cli-gHElfca4SKi1v44yZmrxVA== to 6 log tells you) and Bloop is cancelling the duplication correctly because there are no compile updates from the original compilation that your client is trying to deduplicate. This hints at the compilation of that project being stuck because that only happens when there are no progress updates from the compiler itself. When this happens, the compilation is always cancelled but if the compilation is stucked then the compiler might be ignoring that cancellation. I would check stack sizes and memory limits, as those are the most common reason for stuck compilations.

[tpolecat]

This has happened to me once or twice recently but it seems that brew services restart bloop now unconfuses it, at least temporarily (I no longer need to wipe out the .bloop directory and re-import my project). So at least for me it’s less severe and much less frequent than before, but hasn’t disappeared entirely.