Re: Re: problem starting queue on ia64 + rhel 3AS

[Amol Kulkarni]

 
yes i added 1 or 2 messages to see the flow. also the next call is read on file descriptor 2 which i think is failing.
as far as i could understand qmail-send is expecting qmail-lspawn on file descriptor 2. it seems that qmail-lspawn didnt start.

i'll try compiling with -g option.

thanks very much for ur suggestion, and i'll post the outcome. 


On Tue, 13 Dec 2005 Claudio Jeker wrote :
>On Tue, Dec 13, 2005 at 11:13:46AM -0000, Amol  Kulkarni wrote:
> >
> > i'm able to compile the latest qmail-ldap on ia64 + rhel 3AS but the
> > qmail queue does not start. the control files seem to be ok.
> > so i ran foll. strace cmd (after setting /var/qmail/bin in PATH) :
> >
> > strace ./qmail-start "./Maildir/" splogger qmail
> >
> > the output shows a segmentation fault, but i could not go beyond to find
> > out why this is happening. can anyone please help me in debugging this
> > problem?
> >
>
>Normaly it is the best to get a core dump from a qmail-send compiled and
>linked with -g. The you can get a nice backtrace from gdb.
>
> >
> > chdir("queue")                          = 0
> > rt_sigaction(SIGPIPE, {SIG_IGN}, NULL, 8) = 0
> > rt_sigaction(SIGTERM, {0x4000000000015f70, [], 0}, NULL, 8) = 0
> > rt_sigaction(SIGALRM, {0x4000000000015f30, [], 0}, NULL, 8) = 0
> > rt_sigaction(SIGHUP, {0x4000000000015f60, [], 0}, NULL, 8) = 0
> > rt_sigaction(SIGCHLD, {SIG_DFL}, NULL, 8) = 0
> > umask(077)                              = 077
> > open("lock/sendmutex", O_WRONLY|O_NONBLOCK) = 9
> > flock(9, LOCK_EX|LOCK_NB)              = 0
>
>This is is the start of qmail-send.
>
> > write(0, "info: opened mutex\n", 19)    = 19
>
>This message was added by you (at least it is not in my source)
>
> > fstat(1, {st_mode=S_IFIFO|0600, st_size=0, ...}) = 0
> > mmap(NULL, 65536, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x2000000000040000
> > write(0, "info: next channel \n", 20)  = 20
>
>No idea where that is comming from.
> From the program flow the next syscall should be read() ???
>
>--
>:wq Claudio