MacOS: TB cluster memory allocator error

[kassane]

https://github.com/kassane/tigerbeetle-cpp/actions/runs/6212521694/job/16862738985#step:6:1

[kassane]

Upstream crash!!

Ref.: tigerbeetle/tigerbeetle#1037

[kassane]

Again: https://github.com/kassane/tigerbeetle-cpp/actions/runs/6342352462/job/17228294615

[kassane]

after #21 and tigerbeetle/tigerbeetle#1209 on MacOS persist error

Ref.: https://github.com/kassane/tigerbeetle-cpp/actions/runs/6472993099/job/17574753331

Starting replica 0

running client...
info(message_bus): connected to replica 0
/Users/runner/work/tigerbeetle-cpp/tigerbeetle-cpp/scripts/runner.sh: line 31:  3258 Segmentation fault: 11  $1
Error running with tigerbeetle
info(io): opening "0_0.tigerbeetle"...
info(main): 0: Allocated 5347MB in 16 regions during replica init (Grid Cache: 1024MB)
info(main): 0: cluster=0: listening on 127.0.0.1:3001
info(message_bus): connection from client 123649564302046689183064190193078852059
info(message_bus): peer performed an orderly shutdown: message_bus.MessageBusType(.replica).Connection.Peer{ .client = 123649564302046689183064190193078852059 }
make[3]: *** [run_with_tb] Error 139
make[2]: *** [CMakeFiles/run_with_tb.dir/all] Error 2
make[1]: *** [CMakeFiles/run_with_tb.dir/rule] Error 2
make: *** [run_with_tb] Error 2

cc: @batiati

[batiati]

It does not seem to be related to #16 (comment) since it's not in the format phase, neither #16 (comment).

The log message info(message_bus): peer performed an orderly shutdown shows that the client was disconnected, so it's not likely a server-side error.

Also runner.sh: line 31 is where the sample is invoked by the CI, so I suggest you try to obtain some stack trace information running this specific piece of code, so we can inspect the exact location of the segfault.

[kassane]

It does not seem to be related to #16 (comment) since it's not in the format phase, neither #16 (comment).

Oh! It makes sense, really. This new bug isn't correlating although it runs on MacOS. I'll separate this issue.

The log message info(message_bus): peer performed an orderly shutdown shows that the client was disconnected, so it's not likely a server-side error.

Also runner.sh: line 31 is where the sample is invoked by the CI, so I suggest you try to obtain some stack trace information running this specific piece of code, so we can inspect the exact location of the segfault.

This line refers exactly to running the client in runner.sh.

tigerbeetle-cpp/scripts/runner.sh

Lines 30 to 32 in a3d34ce

	echo "running client..."
	$1
	echo ""

But on my tests using C sample client, this error not persists:
https://github.com/kassane/tigerbeetle-cmake/actions/runs/6485996300