In message <1235949090.4257.5.camel@xxxxxxxxxxxxxxxxxx>, Nathan Scott writes:
>On Fri, 2009-02-27 at 11:14 -0600, Scott Emery wrote:
>> If you hand me some recipes, I might be able to use the original
>> machine to help. a verbose dump of the stack... any particular variables
>> you want to look at?
>
>First thing to check is the source line which caused SIGSEGV.
>I don't understand how "client[i].fd = fd" could cause it, as
>we wrote into "client[i].addr" twenty three lines above... it
>seems as if that line _should_ have caused a SIGSEGV first.
>
>Would be handy to know things like how many clients, contents
>and addresses of the "client" array (and is client[i] near a
>page boundary), did we just realloc this in NewClient(), etc.
>
>cheers.
>
>--
>Nathan
>
The pmcd is probably not built debug with symbols. I wonder if
that is messing up the stack trace and all. I'm going to grab the
pcp-2.7.8 release when it is avialable and build it debug, install
and wait for another instance. The current build seems to fail pretty
often, so it may continue to fail with a debug build.
|