Re: Running valgrind and some asserts against squid-3.0-cvs

On Saturday 09 November 2002 12.08, Reuben Farrelly wrote:

> 1. Appears in 2.5 also, possibly harmless (Henrik?) on boot:
>
> ==19169== Conditional jump or move depends on uninitialised
> value(s) ==19169== at 0x420BACDB: (within
> /lib/i686/libc-2.3.1.so) ==19169== by 0x420BA3E5: (within
> /lib/i686/libc-2.3.1.so) ==19169== by 0x808C507:
> mimeGetEntry(char const*, int) (mime.cc:214) ==19169== by
> 0x808C711: mimeGetContentType (mime.cc:289) (valgrind summary shows
> 1939 times at completion)

Harmless.

> 2,3,4 Appear when first object retrieved from the cache however
> valgrind reports these ones as occurring 245868, 281859 and 380243
> times respectively :
>
> ==18835== Conditional jump or move depends on uninitialised
> value(s) ==18835== at 0x420BACDB: (within
> /lib/i686/libc-2.3.1.so) ==18835== by 0x420BA3E5: (within
> /lib/i686/libc-2.3.1.so) ==18835== by 0x804E00A: aclMatchRegex
> (acl.cc:1066)

Also harmless from what I can tell.

> 5. Going to http://windowsupdate.microsoft.com/ has a few
> times today triggered an assert which kills off valgrind as well as
> squid (ugh):
>
> 2002/11/09 19:47:17| clientReplyContext::saveState: saving store
> context 2002/11/09 19:47:48| assertion failed: ssl.cc:228: "len ==
> (size_t)sslState->client.len"
> Aborted
> [root@tornado squid]#

Most likely a Squid bug, triggered by the timing difference from
runing Squid with the overhead of valgrind CPU emulation.. (valgrind
emulates a X86 CPU to be able to do detailed memory debugging)

> 7. I have gotten this a few times upon sending a TERM signal
> to squid. I haven't seen this one before...this seems to be
> reproduceable(ish) on shutdown as squid sits for possibly about 30s
> before this error comes up.
>
> 2002/11/09 20:18:09| FD 15 Closing SNMP socket
> 2002/11/09 20:18:09| assertion failed: comm.cc:1342:
> "fdc_table[fd].active == 1"
> 2002/11/09 20:18:09| Closing unlinkd pipe on FD 10
> 2002/11/09 20:18:09| aioSync: flushing pending I/O operations
[..]> 2002/11/09 20:18:09| aioSync: flushing pending I/O operations
> 2002/11/09 20:18:09| aioSync: done
> 2002/11/09 20:18:09| assertion failed: disk.cc:92: "F->flags.open"
> 2002/11/09 20:18:09| assertion failed: comm_poll.cc:134:
> "F->flags.open" 2002/11/09 20:18:09| assertion failed:
> comm_poll.cc:134: "F->flags.open"
>
> Program exited normally.
> (gdb) bt

To reliably trap such errors in GDB or core files, start Squid with
the -C option..

Regards
Henrik
Received on Sat Nov 09 2002 - 05:06:48 MST