Wireshark-bugs: [Wireshark-bugs] [Bug 12656] Buildbot crash output: fuzz-2016-07-24-1421.pcap

Date: Sat, 13 Aug 2016 22:20:57 +0000

Comment # 31 on bug 12656 from
(In reply to Guy Harris from comment #30)
> For what it's worth, addr_resolv_cleanup(), called from epan_cleanup(),
> called before TShark - and other programs - exit, calls
> eth_name_lookup_cleanup(), which does
> 
>     wmem_free(NULL, manuf_hashtable);
>     manuf_hashtable = NULL;
> 
> but doesn't free anything in the hash table in question, so that might
> render the manufacturer-OID-to-name table entries "definitely unreachable".
> 
> Should we allocate those entries in epan scope, so they're freed-in-bulk by
> epan_cleanup()?

If I #if 0 out all of eth_name_lookup_cleanup(), and run tshark under valgrind
with that capture, it reports

==28661==
==28661== HEAP SUMMARY:
==28661==     in use at exit: 3,159,806 bytes in 41,121 blocks
==28661==   total heap usage: 122,091 allocs, 80,970 frees, 47,139,021 bytes
al$
==28661==
==28661== LEAK SUMMARY:
==28661==    definitely lost: 12,962 bytes in 207 blocks
==28661==    indirectly lost: 1,059,082 bytes in 30,263 blocks
==28661==      possibly lost: 1,172,867 bytes in 9,249 blocks
==28661==    still reachable: 269,464 bytes in 1,045 blocks
==28661==         suppressed: 645,431 bytes in 357 blocks
==28661== Rerun with --leak-check=full to see details of leaked memory
==28661==
==28661== For counts of detected and suppressed errors, rerun with: -v
==28661== Use --track-origins=yes to see where uninitialised values come from
==28661== ERROR SUMMARY: 20 errors from 4 contexts (suppressed: 0 from 0)


which is a lot less "definitely lost" - an a lot more "indirectly lost" and
"possibly lost" (admittedly, with a later version to TShark).


You are receiving this mail because:
  • You are watching all bug changes.