Re: [Monetdb-developers] [Monetdb-sql-checkins] sql/src/storage store.mx, Feb2009, 1.25.2.1, 1.25.2.2
On 2009-04-22 10:28, Niels Nes wrote:
Update of /cvsroot/monetdb/sql/src/storage In directory 23jxhf1.ch3.sourceforge.com:/tmp/cvs-serv10533/src/storage
Modified Files: Tag: Feb2009 store.mx Log Message: fixing bad concurrency bug
Bad enough to rebuild the release candidates?
1) we lost updates Problem was in the handling of the update bat, ie we need to always keep updates in the update bat. But already apply asap.
2) we got extra inserts (the hard problem) We incorrectly kept using an 'empty' bat. These are bats which should stay empty and are shared. But we incorrectly inserted data in it instead of replacing it with a new bat with data.
U store.mx Index: store.mx =================================================================== RCS file: /cvsroot/monetdb/sql/src/storage/store.mx,v retrieving revision 1.25.2.1 retrieving revision 1.25.2.2 diff -u -d -r1.25.2.1 -r1.25.2.2 --- store.mx 27 Feb 2009 22:23:00 -0000 1.25.2.1 +++ store.mx 22 Apr 2009 08:28:22 -0000 1.25.2.2 @@ -1850,8 +1850,8 @@
t->name = NULL; if (bs_debug) - fprintf(stderr, "trans init (%d,%d)\n", - t->stime, t->schema_number ); + fprintf(stderr, "trans (%p) init (%d,%d)\n", + t, t->stime, t->schema_number ); return t; }
@@ -2710,13 +2710,12 @@ sql_schema *s = n->data; sql_schema *os;
- if (isTempSchema(s)) continue;
os = find_sql_schema(tr->parent, s->base.name); if (os/* || (s->base.wtime == 0 && s->base.rtime == 0)*/) { - if (!validate_tables(s, os)) + if (!validate_tables(s, os)) return 0; } }
------------------------------------------------------------------------------ Stay on top of everything new and different, both inside and around Java (TM) technology - register by April 22, and save $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. 300 plus technical and hands-on sessions. Register today. Use priority code J9JMT32. http://p.sf.net/sfu/p _______________________________________________ Monetdb-sql-checkins mailing list Monetdb-sql-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-sql-checkins
-- Sjoerd Mullender
On 22-04-2009 10:35:58 +0200, Sjoerd Mullender wrote:
On 2009-04-22 10:28, Niels Nes wrote:
Update of /cvsroot/monetdb/sql/src/storage In directory 23jxhf1.ch3.sourceforge.com:/tmp/cvs-serv10533/src/storage
Modified Files: Tag: Feb2009 store.mx Log Message: fixing bad concurrency bug
Bad enough to rebuild the release candidates?
There is at least one person using a gumstix who is blocked by this bug.
On Wed, Apr 22, 2009 at 10:35:58AM +0200, Sjoerd Mullender wrote:
On 2009-04-22 10:28, Niels Nes wrote:
Update of /cvsroot/monetdb/sql/src/storage In directory 23jxhf1.ch3.sourceforge.com:/tmp/cvs-serv10533/src/storage
Modified Files: Tag: Feb2009 store.mx Log Message: fixing bad concurrency bug
Bad enough to rebuild the release candidates?
Probably not related to Niels' changes,, but possibly relevant for the release candidates: Tonight's stable testing shows very bad behaviour (segfaults / double frees) with many M5 & SQL tests on Linux: http://monetdb.cwi.nl/testing/projects/monetdb/Stable/MonetDB5/.mTests103/in... http://monetdb.cwi.nl/testing/projects/monetdb/Stable/sql/.mTests103/index_s... Though yet unconfirmed (working on it), I suspect =================================================================== 2009/04/20 - mlkersten: MonetDB5/src/mal/mal_session.mx,1.175.2.13 sql/src/backends/monet5/sql_scenario.mx,1.334.2.3 The handling of the global environment stack was not properly done upon exit of a session. This led to the situation that variables all appeared uninitialized. =================================================================== Due to a buggy checkin by me, there was no Stable testing yesterday, but yesterday's Current testing showed similar problems (AFAIK, the above changes were already propagated and hence included in yesterday Current testing). Stefan
1) we lost updates Problem was in the handling of the update bat, ie we need to always keep updates in the update bat. But already apply asap.
2) we got extra inserts (the hard problem) We incorrectly kept using an 'empty' bat. These are bats which should stay empty and are shared. But we incorrectly inserted data in it instead of replacing it with a new bat with data.
U store.mx Index: store.mx =================================================================== RCS file: /cvsroot/monetdb/sql/src/storage/store.mx,v retrieving revision 1.25.2.1 retrieving revision 1.25.2.2 diff -u -d -r1.25.2.1 -r1.25.2.2 --- store.mx 27 Feb 2009 22:23:00 -0000 1.25.2.1 +++ store.mx 22 Apr 2009 08:28:22 -0000 1.25.2.2 @@ -1850,8 +1850,8 @@
t->name = NULL; if (bs_debug) - fprintf(stderr, "trans init (%d,%d)\n", - t->stime, t->schema_number ); + fprintf(stderr, "trans (%p) init (%d,%d)\n", + t, t->stime, t->schema_number ); return t; }
@@ -2710,13 +2710,12 @@ sql_schema *s = n->data; sql_schema *os;
- if (isTempSchema(s)) continue;
os = find_sql_schema(tr->parent, s->base.name); if (os/* || (s->base.wtime == 0 && s->base.rtime == 0)*/) { - if (!validate_tables(s, os)) + if (!validate_tables(s, os)) return 0; } }
------------------------------------------------------------------------------ Stay on top of everything new and different, both inside and around Java (TM) technology - register by April 22, and save $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. 300 plus technical and hands-on sessions. Register today. Use priority code J9JMT32. http://p.sf.net/sfu/p _______________________________________________ Monetdb-sql-checkins mailing list Monetdb-sql-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-sql-checkins
-- Sjoerd Mullender
------------------------------------------------------------------------------ Stay on top of everything new and different, both inside and around Java (TM) technology - register by April 22, and save $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. 300 plus technical and hands-on sessions. Register today. Use priority code J9JMT32. http://p.sf.net/sfu/p _______________________________________________ Monetdb-sql-checkins mailing list Monetdb-sql-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-sql-checkins
-- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | | CWI, P.O.Box 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4312 |
On 2009-04-22 10:45, Stefan Manegold wrote:
On Wed, Apr 22, 2009 at 10:35:58AM +0200, Sjoerd Mullender wrote:
On 2009-04-22 10:28, Niels Nes wrote:
Update of /cvsroot/monetdb/sql/src/storage In directory 23jxhf1.ch3.sourceforge.com:/tmp/cvs-serv10533/src/storage
Modified Files: Tag: Feb2009 store.mx Log Message: fixing bad concurrency bug Bad enough to rebuild the release candidates?
Probably not related to Niels' changes,, but possibly relevant for the release candidates:
Tonight's stable testing shows very bad behaviour (segfaults / double frees) with many M5 & SQL tests on Linux: http://monetdb.cwi.nl/testing/projects/monetdb/Stable/MonetDB5/.mTests103/in... http://monetdb.cwi.nl/testing/projects/monetdb/Stable/sql/.mTests103/index_s...
I saw that. It didn't make me dance with joy.
Though yet unconfirmed (working on it), I suspect =================================================================== 2009/04/20 - mlkersten: MonetDB5/src/mal/mal_session.mx,1.175.2.13 sql/src/backends/monet5/sql_scenario.mx,1.334.2.3 The handling of the global environment stack was not properly done upon exit of a session. This led to the situation that variables all appeared uninitialized. ===================================================================
Due to a buggy checkin by me, there was no Stable testing yesterday, but yesterday's Current testing showed similar problems (AFAIK, the above changes were already propagated and hence included in yesterday Current testing).
Stefan
1) we lost updates Problem was in the handling of the update bat, ie we need to always keep updates in the update bat. But already apply asap.
2) we got extra inserts (the hard problem) We incorrectly kept using an 'empty' bat. These are bats which should stay empty and are shared. But we incorrectly inserted data in it instead of replacing it with a new bat with data.
U store.mx Index: store.mx =================================================================== RCS file: /cvsroot/monetdb/sql/src/storage/store.mx,v retrieving revision 1.25.2.1 retrieving revision 1.25.2.2 diff -u -d -r1.25.2.1 -r1.25.2.2 --- store.mx 27 Feb 2009 22:23:00 -0000 1.25.2.1 +++ store.mx 22 Apr 2009 08:28:22 -0000 1.25.2.2 @@ -1850,8 +1850,8 @@
t->name = NULL; if (bs_debug) - fprintf(stderr, "trans init (%d,%d)\n", - t->stime, t->schema_number ); + fprintf(stderr, "trans (%p) init (%d,%d)\n", + t, t->stime, t->schema_number ); return t; }
@@ -2710,13 +2710,12 @@ sql_schema *s = n->data; sql_schema *os;
- if (isTempSchema(s)) continue;
os = find_sql_schema(tr->parent, s->base.name); if (os/* || (s->base.wtime == 0 && s->base.rtime == 0)*/) { - if (!validate_tables(s, os)) + if (!validate_tables(s, os)) return 0; } }
------------------------------------------------------------------------------ Stay on top of everything new and different, both inside and around Java (TM) technology - register by April 22, and save $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. 300 plus technical and hands-on sessions. Register today. Use priority code J9JMT32. http://p.sf.net/sfu/p _______________________________________________ Monetdb-sql-checkins mailing list Monetdb-sql-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-sql-checkins
-- Sjoerd Mullender
------------------------------------------------------------------------------ Stay on top of everything new and different, both inside and around Java (TM) technology - register by April 22, and save $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. 300 plus technical and hands-on sessions. Register today. Use priority code J9JMT32. http://p.sf.net/sfu/p _______________________________________________ Monetdb-sql-checkins mailing list Monetdb-sql-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-sql-checkins
-- Sjoerd Mullender
On Wed, Apr 22, 2009 at 10:45:33AM +0200, Stefan Manegold wrote:
On Wed, Apr 22, 2009 at 10:35:58AM +0200, Sjoerd Mullender wrote:
On 2009-04-22 10:28, Niels Nes wrote:
Update of /cvsroot/monetdb/sql/src/storage In directory 23jxhf1.ch3.sourceforge.com:/tmp/cvs-serv10533/src/storage
Modified Files: Tag: Feb2009 store.mx Log Message: fixing bad concurrency bug
Bad enough to rebuild the release candidates?
Probably not related to Niels' changes,, but possibly relevant for the release candidates:
Tonight's stable testing shows very bad behaviour (segfaults / double frees) with many M5 & SQL tests on Linux: http://monetdb.cwi.nl/testing/projects/monetdb/Stable/MonetDB5/.mTests103/in... http://monetdb.cwi.nl/testing/projects/monetdb/Stable/sql/.mTests103/index_s...
Though yet unconfirmed (working on it), I suspect =================================================================== 2009/04/20 - mlkersten: MonetDB5/src/mal/mal_session.mx,1.175.2.13 sql/src/backends/monet5/sql_scenario.mx,1.334.2.3 The handling of the global environment stack was not properly done upon exit of a session. This led to the situation that variables all appeared uninitialized. ===================================================================
Indeed, on my Fedora 10 desktop, these two checkins (didn't test them separately) make the difference between (before) MonetDB5: 2 out of 467 tests produced SIGNIFICANTLY different output sql : 34 out of 757 tests produced SIGNIFICANTLY different output and (after) MonetDB5: 5 out of 467 tests produced SIGNIFICANTLY different output sql : 97 out of 757 tests produced SIGNIFICANTLY different output The extra failing tests expericne segfaults and/or double frees ... ... further analysis pending ... Stefan
Due to a buggy checkin by me, there was no Stable testing yesterday, but yesterday's Current testing showed similar problems (AFAIK, the above changes were already propagated and hence included in yesterday Current testing).
Stefan
1) we lost updates Problem was in the handling of the update bat, ie we need to always keep updates in the update bat. But already apply asap.
2) we got extra inserts (the hard problem) We incorrectly kept using an 'empty' bat. These are bats which should stay empty and are shared. But we incorrectly inserted data in it instead of replacing it with a new bat with data.
U store.mx Index: store.mx =================================================================== RCS file: /cvsroot/monetdb/sql/src/storage/store.mx,v retrieving revision 1.25.2.1 retrieving revision 1.25.2.2 diff -u -d -r1.25.2.1 -r1.25.2.2 --- store.mx 27 Feb 2009 22:23:00 -0000 1.25.2.1 +++ store.mx 22 Apr 2009 08:28:22 -0000 1.25.2.2 @@ -1850,8 +1850,8 @@
t->name = NULL; if (bs_debug) - fprintf(stderr, "trans init (%d,%d)\n", - t->stime, t->schema_number ); + fprintf(stderr, "trans (%p) init (%d,%d)\n", + t, t->stime, t->schema_number ); return t; }
@@ -2710,13 +2710,12 @@ sql_schema *s = n->data; sql_schema *os;
- if (isTempSchema(s)) continue;
os = find_sql_schema(tr->parent, s->base.name); if (os/* || (s->base.wtime == 0 && s->base.rtime == 0)*/) { - if (!validate_tables(s, os)) + if (!validate_tables(s, os)) return 0; } }
------------------------------------------------------------------------------ Stay on top of everything new and different, both inside and around Java (TM) technology - register by April 22, and save $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. 300 plus technical and hands-on sessions. Register today. Use priority code J9JMT32. http://p.sf.net/sfu/p _______________________________________________ Monetdb-sql-checkins mailing list Monetdb-sql-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-sql-checkins
-- Sjoerd Mullender
------------------------------------------------------------------------------ Stay on top of everything new and different, both inside and around Java (TM) technology - register by April 22, and save $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. 300 plus technical and hands-on sessions. Register today. Use priority code J9JMT32. http://p.sf.net/sfu/p _______________________________________________ Monetdb-sql-checkins mailing list Monetdb-sql-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-sql-checkins
-- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | | CWI, P.O.Box 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4312 |
------------------------------------------------------------------------------ Stay on top of everything new and different, both inside and around Java (TM) technology - register by April 22, and save $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. 300 plus technical and hands-on sessions. Register today. Use priority code J9JMT32. http://p.sf.net/sfu/p _______________________________________________ Monetdb-developers mailing list Monetdb-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-developers
-- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | | CWI, P.O.Box 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4312 |
participants (3)
-
Fabian Groffen -
Sjoerd Mullender -
Stefan Manegold