job-ingest: handle fluid_init() failure

Problem: if fluid_init() fails, job-ingest could issue
duplicate/invalid job IDs.

When fluid_init() fails, job-ingest logs an error but continues
on.  Since the "fluid_generator" is just a 64-bit integer, the
fluids generated after that might be the same as those issued
on rank 0, and may start from an incorrect clock offset.

fluid_init() is expected to fail on rank > 16383.
It could also fail in the unlikely event that clock_gettime() fails.

When fluid_init() fails, abort the module.

Author: garlick

src/modules/job-ingest/job-ingest.c

@@ -854,6 +854,7 @@ int mod_main (flux_t *h, int argc, char **argv)
         if (fluid_init (&ctx.gen, 0, fluid_get_timestamp (max_jobid) + 1) < 0) {
             flux_log (h, LOG_ERR, "fluid_init failed");
             errno = EINVAL;
+            goto done;
         }
     }
     else {
@@ -873,12 +874,10 @@ int mod_main (flux_t *h, int argc, char **argv)
             goto done;
         }
         flux_future_destroy (f);
-        /* fluid_init() will fail on rank > 16K.
-         * Just skip loading the job module on those ranks.
-         */
         if (fluid_init (&ctx.gen, rank, timestamp) < 0) {
             flux_log (h, LOG_ERR, "fluid_init failed");
             errno = EINVAL;
+            goto done;
         }
     }
     flux_log (h, LOG_DEBUG, "fluid ts=%jums", (uint64_t)ctx.gen.timestamp);