Long GC stalls

[singpolyma]

Working with haxe on ios and I'm sometimes seeing some quite long GC stalls. For example I just did a run and I see single calls to MarkAll taking sometimes 100-200 ms which is enough to impact UI rendering. Is there anything I can do to help with this? I tried both generational and moving GC just to try them, but both of those just crash the app.

[hughsando]

What framework are you using? It could be that one thread is preventing the GC from starting - the GC requires that all active threads acknowledge the GC cycle before it starts. Usually, a thread would enter a "GC free zone" before it does some blocking operation. 100-200ms is an in-between time - it is possible it is thread related, or maybe there is just a lot of marking to be done.
You could try uncommenting the "#define PROFILE_COLLECT" in Immix.cpp, which should log some stats.
You could also investigate what all the threads are doing in one of these stalls, it may give a clue - say, from breakpointing "Collect" in Immix.cpp.

[singpolyma]

hmm.. I am using a lot of sqlite in another thread, and I don't see anything in https://github.com/HaxeFoundation/hxcpp/blob/master/src/hx/libs/sqlite/Sqlite.cpp that does gc free zone. maybe that's related...

[singpolyma]

Seems this was the issue. fixed as part of #1132