mirror/hdf5
2
0
Fork 0
mirror of https://github.com/HDFGroup/hdf5.git synced 2025-03-25 17:00:45 +08:00
Code Issues Packages Projects Releases Wiki Activity

[svn-r19959] Bug: the VRFY macro, for some reason, called MPI_Finalize() when it encountered

Browse Source 
error and wanted to exit the test program. This was not good since if only a
subset of processes called MPI_Finalize(), the other processes will likely
hang.  That happened in AIX that it would waited till the alarm signal to kill
the processes. Definitely a waste of time.
Solution: Changed it to call MPI_Abort.

That showed another problem. HDF5 has setup atexit post-process to try to close
unclose objects, release resources, etc.  But if the MPI processes have
encountered an error and has been aborted, it is not likely any more MPI calls
can function properly. E.g., it would attempt to free some communicators in
the HDF5 MPIO file handle. It would again hang.
Solution: need to call H5dont_atexit() to disable any atexit post-processing.
This must be done early, like before calling H5open. This is added to each
parallel test main program.

testphdf5.h:
    Changed macros VRFY and MESG. Added comments too.

testphdf5.c:
t_mpi.c:
t_cache.c:
t_shapesame.c:
    Added H5dont_atexit.

Tested: h5committest.
This commit is contained in:
Albert Cheng 2011-01-16 04:44:22 -05:00
parent 1ed956f2cb
commit e6622c86a8
5 changed files with 57 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
testpar/t_cache.c
View File

@ -7236,6 +7236,14 @@ main(int argc, char **argv)
world_server_mpi_rank = mpi_size - 1;
world_mpi_comm = MPI_COMM_WORLD;
/* Attempt to turn off atexit post processing so that in case errors
* happen during the test and the process is aborted, it will not get
* hang in the atexit post processing in which it may try to make MPI
* calls. By then, MPI calls may not work.
*/
if (H5dont_atexit() < 0){
printf("Failed to turn off atexit processing. Continue.\n", mpi_rank);
};
H5open();
express_test = do_express_test();

8
testpar/t_mpi.c
View File

@ -1137,6 +1137,14 @@ main(int argc, char **argv)
MPI_Comm_size(MPI_COMM_WORLD, &mpi_size);
MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);
/* Attempt to turn off atexit post processing so that in case errors
* happen during the test and the process is aborted, it will not get
* hang in the atexit post processing in which it may try to make MPI
* calls. By then, MPI calls may not work.
*/
if (H5dont_atexit() < 0){
printf("Failed to turn off atexit processing. Continue.\n", mpi_rank);
};
H5open();
if (parse_options(argc, argv) != 0){
if (MAINPROCESS)

9
testpar/t_shapesame.c
View File

@ -4753,6 +4753,15 @@ int main(int argc, char **argv)
printf("PHDF5 TESTS START\n");
printf("===================================\n");
}
/* Attempt to turn off atexit post processing so that in case errors
* happen during the test and the process is aborted, it will not get
* hang in the atexit post processing in which it may try to make MPI
* calls. By then, MPI calls may not work.
*/
if (H5dont_atexit() < 0){
printf("Failed to turn off atexit processing. Continue.\n", mpi_rank);
};
H5open();
h5_show_hostname();

37
testpar/testpar.h
View File

@ -26,25 +26,36 @@
/* Define some handy debugging shorthands, routines, ... */
/* debugging tools */
#define MESG(x) \
if (VERBOSE_MED) printf("%s\n", x); \
/* Print message mesg if verbose level is at least medium and
* mesg is not an empty string.
*/
#define MESG(mesg) \
if (VERBOSE_MED && *mesg != '\0') \
printf("%s\n", mesg)
/*
* VRFY: Verify if the condition val is true.
* If it is true, then call MESG to print mesg, depending on the verbose
* level.
* If val is not true, it prints error messages and if the verbose
* level is lower than medium, it calls MPI_Abort to abort the program.
* If verbose level is at least medium, it will not abort.
* This will allow program to continue and can be used for debugging.
* (The "do {...} while(0)" is to group all the statements as one unit.)
*/
#define VRFY(val, mesg) do { \
if (val) { \
if (*mesg != '\0') { \
MESG(mesg); \
} \
MESG(mesg); \
} else { \
printf("Proc %d: ", mpi_rank); \
printf("*** PHDF5 ERROR ***\n"); \
printf(" Assertion (%s) failed at line %4d in %s\n", \
printf("*** Parallel ERROR ***\n"); \
printf(" VRFY (%s) failed at line %4d in %s\n", \
mesg, (int)__LINE__, __FILE__); \
++nerrors; \
fflush(stdout); \
if (!VERBOSE_MED) { \
printf("aborting MPI process\n"); \
MPI_Finalize(); \
exit(nerrors); \
if (!VERBOSE_MED) { \
printf("aborting MPI processes\n"); \
MPI_Abort(MPI_COMM_WORLD, 1); \
} \
} \
} while(0)
@ -56,9 +67,7 @@
*/
#define INFO(val, mesg) do { \
if (val) { \
if (*mesg != '\0') { \
MESG(mesg); \
} \
MESG(mesg); \
} else { \
printf("Proc %d: ", mpi_rank); \
printf("*** PHDF5 REMARK (not an error) ***\n"); \

9
testpar/testphdf5.c
View File

@ -335,6 +335,15 @@ int main(int argc, char **argv)
printf("PHDF5 TESTS START\n");
printf("===================================\n");
}
/* Attempt to turn off atexit post processing so that in case errors
* happen during the test and the process is aborted, it will not get
* hang in the atexit post processing in which it may try to make MPI
* calls. By then, MPI calls may not work.
*/
if (H5dont_atexit() < 0){
printf("Failed to turn off atexit processing. Continue.\n", mpi_rank);
};
H5open();
h5_show_hostname();