Re: again and again and again

[David Lawrence <davidl@jlab.org>]

Hi Mark,

    How sure are you that the infinite loop is originating in that 
thread? In particular, that code is from the "main" thread which 
basically sleeps and then wakes up to check if everything is done and 
report the rate etc. In other words, it is designed to be an "infinite 
loop" that mostly sleeps. There should be at least 2 other threads 
existing. One is the event reader thread and the other is the event 
processor thread (assuming you're using only 1). Usually, the culprit is 
the event processing loop. That can be identified by a JEventLoop method 
call somewhere in the backtrace or more easily "LaunchThread" at the top 
of the backtrace.

Regards,
-David

Mark M. Ito wrote:
> Folks,
>
> Help! I'm in a loop!
>
> My fitting program occasionally goes into an infinite loop at present. 
> When I run it under the debugger and it has fallen into the loop and I 
> control-C it and backtrace it, it wakes up in JApplication.cc. Here 
> are the statements that are getting repeated:
>
> 957             do{
> 960                     rem.tv_sec = rem.tv_nsec = 0;
> 961                     nanosleep(&req, &rem);
> 962                     if(rem.tv_sec == 0 && rem.tv_nsec == 0){
> 965                             int delta_NEvents = NEvents - 
> GetEventBufferSize() - last_NEvents;
> 966                             avg_NEvents += delta_NEvents>0 ? 
> delta_NEvents:0;
> 967                             avg_time += sleep_time;
> 968                             rate_instantaneous = 
> (double)delta_NEvents/sleep_time;
> 969                             rate_average = 
> (double)avg_NEvents/avg_time;
> 973                     last_NEvents = NEvents - GetEventBufferSize();
> 976                     if(show_ticker && loops.size()>0)PrintRate();
> 978                     if(SIGINT_RECEIVED)Quit();
> 979                     if(SIGINT_RECEIVED>=3)break;
> 983                     pthread_mutex_lock(&app_mutex);
> 986                     double rem_time = (double)rem.tv_sec + 
> (1.0E-9)*(double)rem.tv_nsec;
> 987                     double slept_time = sleep_time - rem_time;
> 988                     for(unsigned int i=0;i<heartbeats.size();i++){
> 989                             double *hb = heartbeats[i];
> 990                             *hb += slept_time;
> 991                             if(monitor_heartbeat && (*hb > 
> (THREAD_TIMEOUT-1.0)+sleep_time)){
> 988                     for(unsigned int i=0;i<heartbeats.size();i++){
> 1042                    pthread_mutex_unlock(&app_mutex);
> 957             do{
>
> and so forth.
>
> Is there any wisdom about what may have caused it? I'm not even sure 
> that the above mentioned loop is the one that is causing problems. 
> Probably some random memory overwrite on my part, but there may be 
> hints for finding it out there, something of which I am unaware.
>
>  -- Mark
>

-- 

------------------------------------------------------------------------
 David Lawrence Ph.D.
 Staff Scientist                 Office: (757)269-5567   [[[  [   [ [       
 Jefferson Lab                   Pager:  (757)584-5567   [  [ [ [ [ [   
 http://www.jlab.org/~davidl     davidl@jlab.org         [[[  [[ [[ [[[
------------------------------------------------------------------------