Transmission-daemon crashes with SEGFAULT in libevent only IF libevent double-free bugfix is NOT applied (sibling ticket for Linux) [this is not a Transmission bug]

[cfpp2p]

Random crashes in transmission-daemon.

crashes in libevent evdns.c

​https://github.com/libevent/libevent/commit/bec50680a31834619e66cb8b97edd6d5b5f15701

I'm currently using compiled libevent 2.0.17 nightly which includes:

@@ -896,7 +896,12 @@ struct deferred_reply_callback {

             addrbuf, sizeof(addrbuf)));

       break;

     default:

-      /* we got a good reply from the nameserver */

+      /* we got a good reply from the nameserver: it is up. */

+      if (req->handle == req->ns->probe_request) {

+        /* Avoid double-free */

+        req->ns->probe_request = NULL;

+      }

+

       nameserver_up(req->ns);

     }

I've only experienced 2 or 3 random crashes in the past few months but I'll update if another similar crash occurs.

[jordan]

cfpp2p, you need to provide a crash report / backtrace in order for the developers to see where Transmission is crashing... please see https://trac.transmissionbt.com/ticket/4768#comment:8 for information on how to generate that backtrace. Thanks!

[cfpp2p]

Jordon,

I almost never get crashes and the two or three crashes I had in libevent evdns.c were months ago, so I will have no backtrace for this #4774 issue.

What I was trying to do with this ticket was present the sibling ticket for #4646 :

But...

For Linux users, I'm tempted to mark this as "invalid" since it's out of our hands and relies on what version of libevent gets bundled by the Linux distros.

So, what's The Right Thing for handling this ticket? There's probably not one... but I'm going to recategorize this ticket for the Mac build and milestone for 2.50.

If someone wants to file a sibling ticket for Linux s.t. I can properly close it as invalid, feel free ;) 

SO, I expected you to close this ticket as invalid, and I was trying to say that if I get the crash again with the libevent patch being applied I would update THIS #4774 ticket...

[livings124]

Please don't create tickets for imaginary issues.

[x190]

Linux users should refer to comment:13 in #4646 for instructions on how to proceed if they are getting the crash detailed in that original ticket.

[cfpp2p]

Please don't create tickets for imaginary issues.

?? Please don't create tickets for imaginary issues. ??

wtf ?????????????????????????????

please don't be such a jerk.........

A great project, but wonderfully conceited developers

I'm sick of this kind of behavior

What great geniuses..., however can't even read plain English or relate to people in a normal social manner.

I worked hard to try to contribute but your ridiculous insulting responses continue.

I was just trying to help.

see you later....

Indications are that devs want cfpp2p to go away, SO

BYE BYE

[livings124]

What just happened? You created a ticket you expected to be closed as invalid, because it was not an issue with Transmission. What did you expect?

[jordan]

cfpp2p: I didn't close this ticket as invalid because your description said that you've already using the double-free bugfix. I thought you were reporting a new crash different from #4646.

If you're seeing a new crash, obviously I'd like to hear more about that, especially a backtrace.

I'm deliberately ignoring comments 3 through 6 :)

[cfpp2p]

Thank you Jordon !

Yes, exactly, as you have stated above, the crashes occurred BEFORE the double-free bugfix. I currently have the double-free bugfix compiled and running in libevent 2.0.17. WITH the double-free bugfix currently applied if I ever see a new crash pertaining to crashes in libevent evdns.c (or otherwise), I will then provide the backtrace and related circumstances leading to the crash.

If you please, go ahead and modify description to reflect what is now clarified. (I can't seem to modify the 'Description')

Too bad about all the confusion, but we've got it right now.