Joel Sherrill commented on a discussion on cpukit/rtems/src/msgqconstruct.c: 
https://gitlab.rtems.org/rtems/rtos/rtems/-/merge_requests/1075#note_143495

 >      is_global = false;
 >    }
 >  
 > -#if 1
 >    /*
 > -   * I am not 100% sure this should be an error.
 > -   * It seems reasonable to create a que with a large max size,
 > -   * and then just send smaller msgs from remote (or all) nodes.
 > +   * A global queue may receive remote sends over MPCI packets.  The queue

This code fragment has moved around so I went back to 4.10 to see who wrote it. 
git blame shows:

`1e1b3e00d1d (Joel Sherrill  1999-05-17 22:52:59 +0000  90) #if 1`

After 27 years, I am more sure this should be an error. It is very likely a 
system configuration mistake to have a message queue set as global that cannot 
send its full length message over the provided MPCI implementation. This would 
now be called a systems integration error. Better to find it out upfront in a 
lab than randomly failing to send some messages.

-- 
View it on GitLab: 
https://gitlab.rtems.org/rtems/rtos/rtems/-/merge_requests/1075#note_143495
You're receiving this email because of your account on gitlab.rtems.org.


_______________________________________________
bugs mailing list
[email protected]
http://lists.rtems.org/mailman/listinfo/bugs

Reply via email to