Re: [Tarantool-patches] [PATCH luajit v2 4/5] build: introduce LUAJIT_USE_ASAN option

[Igor Munkin via Tarantool-patches <tarantool-patches@dev.tarantool.org>]

Sergey,

Thanks for your review! Please consider my answer inline (though, we've
already discussed it offline).

On 25.07.23, Sergey Bronnikov wrote:
> Hello, again
> 
> 
> On 7/21/23 11:12, Igor Munkin wrote:

<snipped>

> 
> I suspect we need set in LUAJIT_USE_SYSMALLOC in cmake/luajit.cmake [1] on
> next LuaJIT bump, right?

Again, all problems come from #3071[1]: lua_State is not closed in scope
of the patched <os.exit> in Tarantool. Hence, leaks occur in ASan/LSan
reports (even with true given as a second parameter of <os.exit>). Until
this issue is not resolved, LUAJIT_USE_SYSMALLOC will report lots of
"false positives" issues (at least, I can't prove that these are real
leaks in LuaJIT/Tarantool). I'll change this when #3071 is resolved (the
same way it's done for LUAJIT_USE_VALGRIND).

Hence, LuaJIT-related ASan checks is relaxed in Tarantool, but we have
strong validation in LuaJIT CI.

> 
> Otherwise Tarantool CI will fail.
> 
> 
> 1. https://github.com/tarantool/tarantool/blob/1cda22aaaf7eb1f0880f4c70f2c79ce27a9604ce/cmake/luajit.cmake#L74-L76
> 
> 
> >   # --- Main source tree ---------------------------------------------------------

[1]: https://github.com/tarantool/tarantool/issues/3071

-- 
Best regards,
IM