Sergey Kaplun
Sergey Kaplun
The same problem occurs on arm64 workflows. The JIT compiler fails to `mmap` mcode area in `LJ_TARGET_JUMPRANGE` (+-27Mb for arm64), traces are not compiled and JIT-related tests fail.
The changeset is merged to master in scope of #9832, backported to 3.0 and 2.11 in scope of #9833 and #9834 respectively.
The changeset is merged to master in scope of #9737, backported to 3.0 and 2.11 in scope of #9738 and #9739 respectively.
The changeset is merged to master in scope of https://github.com/tarantool/tarantool/pull/9938, backported to 3.0 and 2.11 in scope of https://github.com/tarantool/tarantool/pull/9939 and https://github.com/tarantool/tarantool/pull/9940 respectively.
The changeset is merged to master in scope of #10423, backported to 3.1 and 2.11 in scope of #10424 and #10425 respectively.
The changeset is merged to master in scope of #10200, backported to 3.1 and 2.11 in scope of #10201 and #10202 respectively.
Looks like this issue is resolved by LuaJIT/LuaJIT#695 in the upstream.
In the scope of this issue, it would be nice to add the aforementioned benchmarks in our Grafana for comparison, I suppose.
Looks unrelated to Clang bump: I got the same leak report for 2.11 during the LuaJIT bump: https://github.com/tarantool/luajit/actions/runs/11367129005/job/31619134132#step:8:8083
I suppose that the reason is the same as for tarantool/tarantool-qa#233, tarantool/tarantool-qa#276 -- the `cb()` is referenced on the trace: ``` ---- TRACE 22 start 13/2 net_box.lua:703 ---- TRACE 22...