- 30 Apr, 2019 29 commits
-
-
Danny Auble authored
-
Tim Wickberg authored
-
Tim Wickberg authored
Still respect "cray" as the keyword here for now, but add "datawarp" as the documented form instead. Bug 6440.
-
Tim Wickberg authored
-
Danny Auble authored
Since the slurmd will change an older protocol version to the current version we need to make sure if an old srun sends a message we handle things correctly.
-
Danny Auble authored
-
Alejandro Sanchez authored
Otherwise non-FairTree systems will have a decay period delay before the updated value makes it into the sum. Bug 5202.
-
Brian Christiansen authored
-
Brian Christiansen authored
-
Jason Booth authored
Usagefactor matches the documentation and now multiplies TRES time limits and usage. Bug 5435
-
Dineshkumar RAJAGOPAL authored
This is very coarse-grained locking, but as the initial implementation did not anticipate concurrent access this is the safest approach for now. Bug 5638.
-
Tim Wickberg authored
Rather than repeat the #ifdef block through five locations, use a common function to handle this and print a consistent error message if there is an issue. From getrlimit(2): setrlimit() now returns with errno set to EINVAL in places that historically succeeded. It no longer accepts "rlim_cur = RLIM_INFINITY" for RLIM_NOFILE. Use "rlim_cur = min(OPEN_MAX, rlim_max)". (This was causing spurious error messages to appear for sbatch/srun.)
-
Tim Wickberg authored
macOS does not permit the symbol overloading trick that was previously used here. Sending function pointers in as part of slurm_opt_t should work around this.
-
Tim Wickberg authored
-
Tim Wickberg authored
Avoids a dereference of slurmctld_primary, which is a weak symbol, and is not available when the select plugins are loaded in the user commands. (Weak symbols on macOS cannot have local definitions as well - so any reference to them that does not resolve causes the process to crash.)
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Tim Wickberg authored
macOS always has these available without setting -pthread, and ACX_PTHREAD does not work properly on macOS at the moment.
-
Tim Wickberg authored
These should ideally be split off from the task plugin interface since they're tied to a specific implementations. And the core_spec code itself does not really belong directly in slurmd.c.
-
Tim Wickberg authored
There is no equivalent to cpuset_t.
-
Tim Wickberg authored
-
Tim Wickberg authored
Remove initializations - these are not permitted on weak_import'd symbols.
-
Tim Wickberg authored
This was cobbled together from a few different references. All it's doing is forcing a symbol to be created with the slurm_* name, and setting the symbol address to that of the unaliased symbol.
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Doug Jacobsen authored
Bug 3745.
-
- 29 Apr, 2019 11 commits
-
-
Tim Wickberg authored
Using an index value equal to the number of elements puts you one past the end of the array, modify conditional to >= instead. CID 197759.
-
Tim Wickberg authored
Anyone jumping to the cleanup label does not have agent_info_ptr set. CID 197735.
-
Nate Rini authored
Bug 6513
-
Nate Rini authored
Bug 6513
-
Nate Rini authored
Bug 6513
-
Nate Rini authored
Bug 6513
-
Tim Wickberg authored
CID 197758.
-
Tim Wickberg authored
I believe job cannot be NULL here, but the flow through this function is messy, and relies on some external setup that has not been documented with xassert()s. Put this in for now as a stop-gap; this code should be refactored at a later point.
-
Brian Christiansen authored
-
Brian Christiansen authored
-
Tim Wickberg authored
Bug 6632.
-