[debian-hurd-Patches][312017] #552397: openmpi 1.3.3
Patches item #312017, was changed at 18/10/2009 10:05 by Pino Toscano
You can respond by visiting:
https://alioth.debian.org/tracker/?func=detail&atid=410472&aid=312017&group_id=30628
Status: Open
Priority: 3
Submitted By: Pino Toscano (pino-guest)
Assigned to: Nobody (None)
>Summary: #552397: openmpi 1.3.3
Category: None
Group: None
Resolution: None
Initial Comment:
This should fix the various porting issues of openmpi 1.3.3 on GNU/Hurd:
- missing SA_SIGINFO, so a sa_handler is used; show_stackframe_handler() calls show_stackframe() with NULL parameters as that handler checks for non-NULL siginfo_t* and context* already
- some fixed-size char[] for getcwd() to get_current_dir_name() + free()
- some fixed-size char[] for gethostname() to a dynamic realloc() loop
- one gethostname() call commented out, as it seems unused...
- look for mach-o/arch.h instead of mach/mach_host.h to determine whether it is compiling on Darwin
----------------------------------------------------------------------
>Comment By: Pino Toscano (pino-guest)
Date: 26/10/2009 00:42
Message:
Ups my bad, the count is indeed 19, I must have counted the final null byte twice.
Thanks for the review, I sent the updated patch as bug #552397
----------------------------------------------------------------------
Comment By: Samuel Thibault (sthibault)
Date: 25/10/2009 19:33
Message:
Mmm, why + 20?
I would rather use 13 + strlen(jobidstring) + 1 + strlen(orte_process_info.nodename) + 4 + 1.
----------------------------------------------------------------------
Comment By: Pino Toscano (pino-guest)
Date: 25/10/2009 10:41
Message:
Right, added checks for ENAMETOOLONG in gethostname() loops.
Also added a PATH_MAX change in orte/orted/orted_main.c, needed in 1.3.3-2 (it was not in 1.3.3-1).
----------------------------------------------------------------------
Comment By: Samuel Thibault (sthibault)
Date: 24/10/2009 01:10
Message:
In mca_base_component_find.c and ompi_mpi_finalize.c, you should check
that errno is ENAMETOOLONG, because else if some other error happens
(who knows what), you'll end up eating as much memory as possible.
Else, seems fine, please submit after having added the errno check!
----------------------------------------------------------------------
You can respond by visiting:
https://alioth.debian.org/tracker/?func=detail&atid=410472&aid=312017&group_id=30628
Reply to: