[Gaim-devel] gaim_srv_resolve() / resolve() failures and _sigtramp()

[Evan S. <ev...@dr...>]

Sometimes, resolve() apparently crashes the child process made via  
fork() in gaim_srv_resolve().
This not happening would clearly be optimal... but I don't see what's  
wrong with the call to res_query(). (It appears as res_9_nquery() in  
the below stack trace from sample because OS X sends the call there  
directly).  Somehow, res_query() crashes in res_9_nclose()... see below:
-----
261 -[SLGaimCocoaAdapter setStatusID:isActive:arguments:onAccount:]
   261 jabber_login
    261 gaim_srv_resolve
     261 resolve
      261 res_9_nquery
        261 res_nquery_2
          261 res_9_nsend_2
            261 res_9_nclose
              261 res_9_nclose
                261 cerror
                  261 malloc_zone_calloc
                    261 0xfffffffc
                      261 _sigtramp
                        261 CrashHandler_Signal
                          261 -[NSPathStore2  
stringByAppendingPathComponent:]
                            261 +[NSPathStore2  
pathStoreWithCharacters:length:]
                              261 NSAllocateObject
                                261  
_internal_class_createInstanceFromZone
                                  261 malloc_zone_calloc
                                    261 0xfffffffc
                                      261 0xfffffffc
-----
It crashes, but in the case of Adium, a signal handler has been  
installed previously to handle crashes.  The signal handler, which  
uses Cocoa code, ends up stuck in an infinite loop as a result (Cocoa  
frameworks can't be used from a fork()'d process without other magic  
because memory addresses are shared with the parent, I understand).

Would it be reasonable to clear signal handlers in the child process  
before resolve() is called?  If not, how should this be handled?

Thanks,
Evan