This is not correct. I made the exact opposite change in changeset 080dcfcb43e4 in order to fix memory leaks. That worked, but now those leaks are back. Clearly another solution is needed.

On 28/07/16 00:31, Martin Kersten wrote:

Changeset: 25c896ae6a53 for MonetDB URL: http://dev.monetdb.org/hg/MonetDB?cmd=changeset;node=25c896ae6a53 Modified Files: monetdb5/mal/mal_profiler.c Branch: Jun2016 Log Message:

Trace BATs should be reset before the next TRACE call BBPunfix was slipped in, while it should simply be a BATclear.

diffs (36 lines):

diff --git a/monetdb5/mal/mal_profiler.c b/monetdb5/mal/mal_profiler.c --- a/monetdb5/mal/mal_profiler.c +++ b/monetdb5/mal/mal_profiler.c @@ -761,19 +761,19 @@ clearTrace(void) return; /* not initialized */ } /* drop all trace tables */

• BBPunfix(TRACE_id_event->batCacheid);
• BBPunfix(TRACE_id_time->batCacheid);
• BBPunfix(TRACE_id_pc->batCacheid);
• BBPunfix(TRACE_id_thread->batCacheid);
• BBPunfix(TRACE_id_ticks->batCacheid);
• BBPunfix(TRACE_id_rssMB->batCacheid);
• BBPunfix(TRACE_id_tmpspace->batCacheid);
• BBPunfix(TRACE_id_inblock->batCacheid);
• BBPunfix(TRACE_id_oublock->batCacheid);
• BBPunfix(TRACE_id_minflt->batCacheid);
• BBPunfix(TRACE_id_majflt->batCacheid);
• BBPunfix(TRACE_id_nvcsw->batCacheid);
• BBPunfix(TRACE_id_stmt->batCacheid);

• BBPclear(TRACE_id_event->batCacheid);
• BBPclear(TRACE_id_time->batCacheid);
• BBPclear(TRACE_id_pc->batCacheid);
• BBPclear(TRACE_id_thread->batCacheid);
• BBPclear(TRACE_id_ticks->batCacheid);
• BBPclear(TRACE_id_rssMB->batCacheid);
• BBPclear(TRACE_id_tmpspace->batCacheid);
• BBPclear(TRACE_id_inblock->batCacheid);
• BBPclear(TRACE_id_oublock->batCacheid);
• BBPclear(TRACE_id_minflt->batCacheid);
• BBPclear(TRACE_id_majflt->batCacheid);
• BBPclear(TRACE_id_nvcsw->batCacheid);
• BBPclear(TRACE_id_stmt->batCacheid); TRACE_init = 0; MT_lock_unset(&mal_contextLock); initTrace();

---

checkin-list mailing list checkin-list@monetdb.org https://www.monetdb.org/mailman/listinfo/checkin-list