Re: [PATCH 2/2] Rewrite gsync so that it works with remote tasks v2

[Samuel Thibault]

Hello,

Did you actually try with rpctrace?  It seems to be breaking it. For
instance:

#include <pthread.h>
#include <stdio.h>

pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
void *f(void *foo)
{
        pthread_mutex_lock(&mutex);
        sleep(1);
        pthread_mutex_unlock(&mutex);
        return NULL;
}
int main(void) {
        pthread_t t1, t2;
        pthread_create(&t1, NULL, f, NULL);
        pthread_create(&t2, NULL, f, NULL);
        pthread_join(t1, NULL);
        pthread_join(t2, NULL);
        return 0;
}

When running this through rpctrace, rpctrace hangs, with the following
backtrace:

(gdb) bt
#0  0x010999ac in mach_msg_trap ()
    at 
/build/glibc-cU1xjk/glibc-2.24/build-tree/hurd-i386-libc/mach/mach_msg_trap.S:2
#1  0x0109a146 in __mach_msg (msg=0x3003f10, option=1, send_size=64, 
    rcv_size=0, rcv_name=0, timeout=0, notify=0) at msg.c:110
#2  0x0804b812 in trace_and_forward (inp=0x3003f10, outp=0x3001f00)
    at ../../utils/rpctrace.c:1357
#3  0x0105a23c in internal_demuxer (inp=<optimized out>, 
    outheadp=<optimized out>) at ../../libports/manage-one-thread.c:96
#4  0x0109a6fe in __mach_msg_server_timeout (demux=0x3005f8c, max_size=8192, 
    rcv_name=55, option=2048, timeout=0) at msgserver.c:108
#5  0x0105a2e5 in ports_manage_port_operations_one_thread (
    demuxer=0x804b5c0 <trace_and_forward>, timeout=0)
    at ../../libports/manage-one-thread.c:120
#6  0x0804a043 in trace_thread_function (arg=0x8059bc8)
    at ../../utils/rpctrace.c:1382
#7  0x0106cd2f in entry_point (self=0x80562f8, 
    start_routine=0x804a030 <trace_thread_function>, arg=0x8059bc8)
    at ./pthread/pt-create.c:64

AIUI, the problem is that rpctrace uses only one thread to do the
tracing, and since gsync_wait is blocking the RPC, the whole rpctrace
gets blocked.  Unfortunately, that prevents from commiting such change,
since not being able to rpctrace a threaded program is really a concern.

Samuel