[erlang-questions] Tracking memory leaks in NIF/Linkedin drivers with Valgrind (or anything else)

[Zabrane Mickael]

Sverker,

Tried your suppression file (with --show-possibly-lost=no) and got this when starting/stopping a node:

[...]
--4240-- 
--4240-- used_suppression:      1 Early permanent cache aligned ts_event_pool
--4240-- used_suppression:      1 Permanent cache aligned malloc for array of mseg allocators
--4240-- used_suppression:      1 Harmless leak of ErtsThrPrgrData from async threads in exiting emulator
--4240-- used_suppression:      2 No leak; pointer into block
--4240-- used_suppression:      1 Early permanent cache aligned erl_process:aux_work_tmo
--4240-- used_suppression:      1 Prebuilt constant terms in os_info_init (PossiblyLost)
==4240== 
==4240== ERROR SUMMARY: 416 errors from 9 contexts (suppressed: 4 from 4)

Can you please confirm that you get the same "false errors" in your side?

Regards,
Zabrane

On Mar 18, 2013, at 7:09 PM, Sverker Eriksson wrote:

> You don't need the patched version if you run with valgrind option --show-possibly-lost=no.
> This is what we do currently in our daily tests.
> 
> 
> export VALGRIND_MISC_FLAGS="--suppressions=$ERL_TOP/erts/emulator/valgrind/suppress.standard --show-possibly-lost=no"
> 
> /Sverker