I support this proposal 100% ! Andreas On Fri, 11 Feb 2005, Daniel Templeton wrote:
Here's my proposal for the new wait structure:
typedef struct drmaa_job_info_s drmaa_job_info_t;
We then then add the functions:
void drmaa_release_job_info(drmaa_job_info_t* info);
int drmaa_wreason(char *reason, size_t reason_len, drmaa_job_info_t *info, char* error_diagnosis, size_t error_diag_len);
int drmaa_wrusage(drmaa_attr_values_t **rusage, drmaa_job_info_t *info, char* error_diagnosis, size_t error_diag_len);
int drmaa_wjobid(char *jobid, size_t jobid_len, drmaa_job_info_t *info, char* error_diagnosis, size_t error_diag_len);
and change existing functions like this:
int drmaa_wait(const char *job_id, signed long timeout, drmaa_job_info_t **info, char *error_diagnosis, size_t error_diag_len);
int drmaa_wifexited(int *exited, drmaa_job_info_t *info, char* error_diagnosis, size_t error_diag_len);
int drmaa_wexitstatus(int *exit_status, drmaa_job_info_t *info, char* error_diagnosis, size_t error_diag_len);
int drmaa_wifsignaled(int *signaled, drmaa_job_info_t *info, char* error_diagnosis, size_t error_diag_len);
int drmaa_wtermsig(char *signal, size_t signal_len, drmaa_job_info_t *info, char* error_diagnosis, size_t error_diag_len);
int drmaa_wcoredump(int *core_dumped, drmaa_job_info_t *info, char* error_diagnosis, size_t error_diag_len);
int drmaa_wifaborted(int *aborted, drmaa_job_info_t *info, char* error_diagnosis, size_t error_diag_len);
I put the job_id_out and rusage into the new structure because that's how it's handled in the OO specs. If the consensus is that we'd rather keep them as individual parameters, we can do that too. I would argue that the job_id_out definitely belong in the new structure. Barring what the OO specs do, I'm not sure how I feel about rusage. The error codes and function descriptions would obviously mirror those of the peer functions.
Daniel