[bugzilla-daemon@xxxxxxxxxxxxx]

Peter Wu changed bug 13044

What	Removed	Added
CC		peter@lekensteyn.nl

Comment # 16 on bug 13044 from Peter Wu

I'm unable to reproduce the uninitialized variable warning under valgrind 3.12,
gcc 6.2.1 on Arch Linux x86_64. ASAN does also not complain. Manual inspection
suggests that everything should work.

On Ubuntu 16.04 I can also not reproduce it with CMake (both
-DCMAKE_BUILD_TYPE=Debug and -DCMAKE_BUILD_TYPE=DebWithRelInfo) nor autotools
(./configure without options) with commit
ce9405b00b0bc49fef5a21f984251a1e0065580b. The only complaint I get is:

==29022== Memcheck, a memory error detector
==29022== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==29022== Using Valgrind-3.11.0 and LibVEX; rerun with -h for copyright info
==29022== Command: run/tshark -nr /tmp/fuzz-2016-10-25-19751.pcap
==29022== 
==29022== Conditional jump or move depends on uninitialised value(s)
==29022==    at 0xA9B59D6: ws_mempbrk_sse42_compile (ws_mempbrk_sse42.c:69)
==29022==    by 0x726EC02: register_all_protocols (register.c:2725)
==29022==    by 0x73463F7: proto_init (proto.c:534)
==29022==    by 0x7322746: epan_init (epan.c:132)
==29022==    by 0x115E64: main (tshark.c:805)
==29022== 


Is anyone else able to reproduce this valgrind report?

The only reason why this could possibly fail is when the compiler did an
optimization which was not understood by valgrind. opcode is a guint8,
solicited a gboolean (just an int), maybe the compiler "optimized" this by
loading a 64-bit number...

---

      You are receiving this mail because:

• You are watching all bug changes.