Idr Idr - 1 year ago 141
C Question

Possible Memory Leak Valgrind in OSX El Capitan

I'm getting a warning for

possibly lost: 2,064 bytes in 1 blocks
when using Valgrind on OSX Yosemite. Is there a fix to this? I installed valgrind using brew.

Below is an example of how to reproduce

~/cat hello.c
int main() {
return 123;

~/uname -a
Darwin mac.local 15.2.0 Darwin Kernel Version 15.2.0: Fri Nov 13 19:56:56 PST 2015; root:xnu-3248.20.55~2/RELEASE_X86_64 x86_64 i386 MacBookAir6,2 Darwin

~/clang --version
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.2.0
Thread model: posix

~/valgrind --version

~/brew info valgrind
valgrind: stable 3.11.0 (bottled), HEAD
Dynamic analysis tools (memory, debug, profiling)
/usr/local/Cellar/valgrind/3.11.0 (328 files, 46.7M) *
Poured from bottle

~/clang hello.c -o hello.o

~/valgrind --leak-check=full ./hello.o
==7972== Memcheck, a memory error detector
==7972== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==7972== Using Valgrind-3.11.0 and LibVEX; rerun with -h for copyright info
==7972== Command: ./hello.o
==7972== HEAP SUMMARY:
==7972== in use at exit: 22,411 bytes in 187 blocks
==7972== total heap usage: 271 allocs, 84 frees, 28,651 bytes allocated
==7972== 2,064 bytes in 1 blocks are possibly lost in loss record 57 of 62
==7972== at 0x10000817C: malloc_zone_malloc (in /usr/local/Cellar/valgrind/3.11.0/lib/valgrind/
==7972== by 0x1004F3EFD: _objc_copyClassNamesForImage (in /usr/lib/libobjc.A.dylib)
==7972== by 0x1004E7182: protocols() (in /usr/lib/libobjc.A.dylib)
==7972== by 0x1004E7093: readClass(objc_class*, bool, bool) (in /usr/lib/libobjc.A.dylib)
==7972== by 0x1004E4C13: gc_init (in /usr/lib/libobjc.A.dylib)
==7972== by 0x1004EC24E: objc_initializeClassPair_internal(objc_class*, char const*, objc_class*, objc_class*) (in /usr/lib/libobjc.A.dylib)
==7972== by 0x1004F9132: layout_string_create (in /usr/lib/libobjc.A.dylib)
==7972== by 0x1004E783C: realizeClass(objc_class*) (in /usr/lib/libobjc.A.dylib)
==7972== by 0x1004E7300: copySwiftV1MangledName(char const*, bool) (in /usr/lib/libobjc.A.dylib)
==7972== by 0x1004E72E9: copySwiftV1MangledName(char const*, bool) (in /usr/lib/libobjc.A.dylib)
==7972== by 0x1004E72E9: copySwiftV1MangledName(char const*, bool) (in /usr/lib/libobjc.A.dylib)
==7972== by 0x1004E72E9: copySwiftV1MangledName(char const*, bool) (in /usr/lib/libobjc.A.dylib)
==7972== LEAK SUMMARY:
==7972== definitely lost: 0 bytes in 0 blocks
==7972== indirectly lost: 0 bytes in 0 blocks
==7972== possibly lost: 2,064 bytes in 1 blocks
==7972== still reachable: 0 bytes in 0 blocks
==7972== suppressed: 20,347 bytes in 186 blocks
==7972== For counts of detected and suppressed errors, rerun with: -v
==7972== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 17 from 17)

Answer Source

Valgrind is mostly a tool for Linux, and is less supported for OSX. This means that Valgrind will generate a lot of false positives on OSX. If you want to suppress those possibly lost leaks, then add the --gen-suppressions=all (or --gen-suppressions=yes if you want to pick and choose reported leaks one by one) option to your valgrind call. What this will do is print off a chunk of text for each reported memory leak that will look something like this:

   match-leak-kinds: indirect

Copy and paste that, brackets and all, to a file called something like /Users/username/leak1.supp. Feel free to change the <...> to an actual name for your suppression. Then when you call valgrind, if you add a --suppressions=/Users/<username>/leak1.supp option, that memory leak report will be suppressed. To make this easier, you can just put stuff in a ~/.valgrindrc file. This file could look something like


Or if you can just test your code on a Linux machine instead, you won't have to worry about all this ;)


I got a lot of my info from this other SO post

Recommended from our users: Dynamic Network Monitoring from WhatsUp Gold from IPSwitch. Free Download