after MPI-startup.
For this a new mpirun-parameter "mpi_signal" is added, one may specify a
comma-separated list of signals to grab, e.g. mpirun --mca mpi_signal 8,11
will check for SIGFPE and SIGSEGV.
It only finds the first fault (SA_ONESHOT), as after the return the same
fault will occur again.
As printout, the data provided by siginfo_t is printed to STDOUT (yes,
it calls printf ,-]).
Additionally, with glibc, it uses backtrace and backtrace_symbols to
print the calling stack up to the function in which the signal was raised:
(Rank:0) Going to write to RD_ONLY mmaped shared mem
Signal:11 info.si_errno:0(Success) si_code:2(SEGV_ACCERR)
Failing at addr:0x4020c000
[0] func:/home/rusraink/ompi-gcc/lib/libmpi.so.0 [0x40121afe]
[1] func:./t0 [0x42029180]
[2] func:./t0(__libc_start_main+0x95) [0x42017589]
[3] func:./t0(__libc_start_main+0x49) [0x8048691]
This commit was SVN r4170.
from themselves -- but doesn't look valid. For example, it's legal to
have $FC and $F77 to both be ifort (intel fortran compiler).
This commit was SVN r4148.
removing mpga.h include as it breaks in the latest Mellanox VAPI releases (actually, mpga.h
is moved to mpga/mpga.h...yuck!)
This commit was SVN r4146.
Remove all useless labels.
Remove all "unsigned compared to signed" warnings
Chack that the datatypes are correctly committed before any usage of pack/unpack/copy functions (The reason is that
on the commit step we add a fake record at the end of the datatype representation. This record contain a END_LOOP
corresponding to the count used for the convertor. This allow us to remove few tests in the critical path, and to
slighty improuve the performances of these functions).
Remove useless code.
Correct the ompi_ddt_copy_to_same_ddt function (now the exit condition is correctly detected in all cases).
Correct the ompi_ddt_dump_stack function (The stack is already pointing to the current position not to the begining
of the stack).
This commit was SVN r4103.
Note: in bringing this across, I have obeyed the mca-prefix-rule in defining functions and variables. However, I left the name of the component call as "orte_errmgr.xxx" so that any use of the code will directly "merge" with the new RTE branch without changes.
This commit was SVN r4099.
t value from -1 to 0xffffffff. Remove the volatile from the locks definition, as th
ey are useless ie. only the internal integer should be volatile (and that's already the case).
Left something inside for the SM maintainer:
- one warning in ptl_sm.c. For me it look like a bug (line 856). The PTL does not h
ave to signal that the request is completed. That's the PML job. Anyway, the ompi_r
equest_complete is called with a mca_pml_base_send_request_t, and that's not exactl
y how it should work.
- in ptl_sm.c line 122 there is a definition (volatile char** tmp_ptr). As far as I
understand the code it's not the char** who should be volatile, but the pointer to
char ...
This commit was SVN r4024.
