Async memory leak when raising exception on ORC
nitely opened this issue · comments
Esteban C Borsani commented
Description
It seems there is a mem leak when raising exceptions from an async proc.
The command I ran:
$ nim c -d:danger --mm:orc --debugger:native --threads:off -d:useMalloc bug.nim && valgrind --leak-check=full --show-leak-kinds=all ./bug
import std/asyncdispatch
proc err {.async.} =
raise newException(ValueError, "err1")
#discard
proc amain {.async.} =
await sleepAsync(1)
try:
await err()
except ValueError:
discard
proc main() =
waitFor amain()
main()
doAssert not hasPendingOperations()
setGlobalDispatcher(nil)
GC_fullCollect()
echo "ok"
Nim Version
Nim Compiler Version 2.1.1 [Linux: amd64]
Compiled at 2024-05-16
Copyright (c) 2006-2024 by Andreas Rumpf
git hash: 2c85515
active boot switches: -d:release
Current Output
Hint: mm: orc; opt: speed; options: -d:danger
63095 lines; 2.754s; 87.977MiB peakmem; proj: /home/esteban/AtomProjects/crap/bug.nim; out: /home/esteban/AtomProjects/crap/bug [SuccessX]
==145078== Memcheck, a memory error detector
==145078== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==145078== Using Valgrind-3.18.1 and LibVEX; rerun with -h for copyright info
==145078== Command: ./bug
==145078==
ok
==145078==
==145078== HEAP SUMMARY:
==145078== in use at exit: 136 bytes in 2 blocks
==145078== total heap usage: 19 allocs, 17 frees, 113,298 bytes allocated
==145078==
==145078== 56 bytes in 1 blocks are indirectly lost in loss record 1 of 2
==145078== at 0x484DCD3: realloc (in /usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so)
==145078== by 0x10CC5F: _ZN6system11reallocImplE7pointer25range09223372036854775807 (malloc.nim:17)
==145078== by 0x10CC5F: _ZN6system14alignedReallocE7pointer25range0922337203685477580725range0922337203685477580725range09223372036854775807 (memalloc.nim:383)
==145078== by 0x10CC5F: prepareSeqAddUninit (seqs_v2.nim:121)
==145078== by 0x10D697: _ZN6system3addE3varI3seqIN6system15StackTraceEntryEEE4sinkIN6system15StackTraceEntryEE (seqs_v2.nim:161)
==145078== by 0x10D697: raiseExceptionEx (excpt.nim:504)
==145078== by 0x1184E8: amain::amainX20X28AsyncX29_ (asyncmacro.nim:170)
==145078== by 0x117297: amain::amainNimAsyncContinue (asyncmacro.nim:35)
==145078== by 0x1188D0: bug::amain (asyncmacro.nim:293)
==145078== by 0x118ADE: _ZN3bug4mainE (bug.nim:15)
==145078== by 0x118ADE: NimMainModule (bug.nim:17)
==145078== by 0x10A3F5: NimMainInner (asyncfutures.nim:234)
==145078== by 0x10A3F5: NimMain (asyncfutures.nim:245)
==145078== by 0x10A3F5: main (asyncfutures.nim:253)
==145078==
==145078== 136 (80 direct, 56 indirect) bytes in 1 blocks are definitely lost in loss record 2 of 2
==145078== at 0x484DA83: calloc (in /usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so)
==145078== by 0x10CF3A: _ZN6system10alloc0ImplE25range09223372036854775807 (malloc.nim:11)
==145078== by 0x10CF3A: _ZN6system13alignedAlloc0E25range0922337203685477580725range09223372036854775807 (memalloc.nim:353)
==145078== by 0x10CF3A: nimNewObj (arc.nim:96)
==145078== by 0x117507: _ZN3err18errX20X28AsyncX29_E (asyncfutures.nim:211)
==145078== by 0x117507: err::errX20X28AsyncX29_ (asyncmacro.nim:250)
==145078== by 0x116FAA: err::errNimAsyncContinue (asyncmacro.nim:33)
==145078== by 0x118230: bug::err (asyncmacro.nim:293)
==145078== by 0x118434: amain::amainX20X28AsyncX29_ (bug.nim:10)
==145078== by 0x11725A: amain::amainNimAsyncContinue (asyncmacro.nim:33)
==145078== by 0x1188D0: bug::amain (asyncmacro.nim:293)
==145078== by 0x118ADE: _ZN3bug4mainE (bug.nim:15)
==145078== by 0x118ADE: NimMainModule (bug.nim:17)
==145078== by 0x10A3F5: NimMainInner (asyncfutures.nim:234)
==145078== by 0x10A3F5: NimMain (asyncfutures.nim:245)
==145078== by 0x10A3F5: main (asyncfutures.nim:253)
==145078==
==145078== LEAK SUMMARY:
==145078== definitely lost: 80 bytes in 1 blocks
==145078== indirectly lost: 56 bytes in 1 blocks
==145078== possibly lost: 0 bytes in 0 blocks
==145078== still reachable: 0 bytes in 0 blocks
==145078== suppressed: 0 bytes in 0 blocks
==145078==
==145078== For lists of detected and suppressed errors, rerun with: -s
==145078== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
Expected Output
No leaks
Possible Solution
No response
Additional Information
Valgrind does not show leaks when trying in Nim 1.6.20
Juan Carlos commented
Bisect points to commits in this PR: #23612 (or before/after it, bug appears near those commits). 🤔