[
https://issues.apache.org/jira/browse/DISPATCH-1309?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16809806#comment-16809806
]
michael goulish commented on DISPATCH-1309:
-------------------------------------------
Yee hah!
Chuck's comment reminded me – I believe I have also seen crashes *only* when
the console was attached.
Furthermore, I think I have seen crashes .... maybe not *only* but *more often*
when I was *shutting down* a console *while* the network was still running.
I tried that just now – with 1.6 code. I had to start, stop, and restart the
console 11 times, but then it happened. Boom. With this core:
#0 pn_collector_put (collector=0x4242424242424242,
clazz=0x7f0e99c38520 <PN_OBJECT>, context=0x0,
type=type@entry=PN_CONNECTION_WAKE)
at /home/mick/latest/qpid-proton-0.26.0/c/src/core/event.c:134
#1 0x00007f0e99ca6258 in http_thread_run (v=0x2036850)
at /home/mick/latest/qpid-dispatch-1.6.0/src/http-libwebsockets.c:731
#2 0x00007f0e995df50b in start_thread () from /lib64/libpthread.so.0
#3 0x00007f0e988a338f in clone () from /lib64/libc.so.6
Which is one I have seen before.
Now I have *some hope* of getting some kind of baseline, based on number of
crashes per console stop-and-restart, so that I can do some kind of vivisection
of the code.
> Various crashes in 1.6 release
> ------------------------------
>
> Key: DISPATCH-1309
> URL: https://issues.apache.org/jira/browse/DISPATCH-1309
> Project: Qpid Dispatch
> Issue Type: Bug
> Affects Versions: 1.6.0
> Environment: System 'unused':(
> Fedora 5.0.3-200.fc29.x86_64,
> Python 2.7.15,
> Proton master @ eab1f.
> System 'taj':(
> Fedora 4.18.16-200.fc28.x86_64,
> Python 3.6.6,
> Proton master @ 68b38
> Reporter: Chuck Rolke
> Priority: Major
> Attachments: DISPATCH-1309-backtraces.txt,
> DISPATCH-1309-gen_configs_linear.py
>
>
> qpid-dispatch master @ 51244, which is very close to the 1.6 release, has
> various crashes.
> The test network is 12 routers spread over two systems. (Configuration
> generator to be attached.) Four interior routers are in linear arrangement
> with A and C on one system ('unused'), and B and D on the other system
> ('taj'). Each system then attaches four edge routers, one to each interior
> router.
> Running lightweight tests, like proton cpp simple_send and simple_recv to
> ports on INTA and INTB interior routers leads to a crash on INTC. The crashes
> typically look like reuse of structures after they have been freed (addresses
> are 0x9999999999999999). Other crashes hint of general memory corruption
> (crashes in malloc.c).
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
