Bruce Momjian [Thu, 3 Jun 1999 20:44:28 +0000 (20:44 +0000)]
SET TRANSACTION ISOLATION LEVEL SERIALIZABLE;
SET TRANSACTION ISOLATION LEVEL READ COMMITTED;
^^^^ required
Also note that SET TRANSACTION ISOLATION LEVEL change
isolevel for _current_ transaction, in accordance with
standard, not for session (ALTER SESSION is used in Oracle,
but it's not implemented currently).
And I would don't mention SET XACTISOLEVEL TO ...
form at all.
Please update set.sgml - I failed to understand all these
Bruce Momjian [Thu, 3 Jun 1999 20:42:18 +0000 (20:42 +0000)]
Thanks for this, but SET TRANSACTION ISOLATION LEVEL
should be described like SET TIME ZONE: TO is not allowed, as well
as DEFAULT option. There are only two forms:
1. MyProc->xid assignment is moved to GetNewTransactionId so newer
transactions will not assume that MyProc transaction was committed
before snapshot calculations. With old MyProc->xid assignment
(in xact.c:StartTransaction()) there was ability to see the same
row twice (I used gdb for this)!...
2. Assignments of InvalidTransactionId to MyProc->xid and MyProc->xmin
are moved from xact.c:CommitTransaction() to
xact.c:RecordTransactionCommit() - this invalidation must be done
before releasing transaction locks or bad (too high) XmaxRecent value
might be used by vacuum ("ERROR: Child itemid marked as unused"
reported by "Hiroshi Inoue" <Inoue@tpf.co.jp>; once again, gdb
allowed me reproduce this error).
Tom Lane [Thu, 3 Jun 1999 04:01:16 +0000 (04:01 +0000)]
pg_dump failed on any database containing large objects,
because it ignored the LOs themselves but failed to ignore the indexes
on the LOs. How long has this been broken??
Make sure that only one intro is included in the integrated doc.
Multiple intros cause trouble since they have some section elements
(e.g. "y2k.sgml") in common leading to duplicate labels.
Include emacs formatting hints in the intro*.sgml sources.
typedef struct LTAG
{
Oid relId;
Oid dbId;
union
{
BlockNumber blkno;
TransactionId xid;
} objId;
>
> Added:
> /*
> * offnum should be part of objId.tupleId above, but would increase
> * sizeof(LOCKTAG) and so moved here; currently used by userlocks only.
> */
> OffsetNumber offnum;
uint16 lockmethod; /* needed by userlocks */
} LOCKTAG;
gmake clean required...
User locks are ready for 6.5 release...
Tom Lane [Mon, 31 May 1999 23:48:04 +0000 (23:48 +0000)]
Fix some latent bugs in dllist.c (carelessness about setting
all fields that should be set). Add a MoveToFront primitive to speed up
one of the hotspots in SearchSysCache.
Tom Lane [Mon, 31 May 1999 17:01:52 +0000 (17:01 +0000)]
Correct serious bug in hashtable expansion routine: under the
right circumstances it would leave old and new bucket headers pointing to
the same list of records.
Vadim B. Mikheev [Mon, 31 May 1999 01:48:13 +0000 (01:48 +0000)]
Release XactLockTable share lock immediately after this lock is acquired
(no sense to hold it) or we'll be out of lock entries.
Great thanks to Hiroshi Inoue.
Tom Lane [Sun, 30 May 1999 15:17:58 +0000 (15:17 +0000)]
Replace static rcsid[] strings by IDENTIFICATION comments in
file headers, to conform to established Postgres coding style and avoid
warnings from gcc.
Tom Lane [Sun, 30 May 1999 00:06:44 +0000 (00:06 +0000)]
configure.in forgot to do AC_LANG_C to reselect C-based tests
after checking for presence of C++ compiler. Odd we hadn't seen any
reports of problems before...
Vadim B. Mikheev [Sat, 29 May 1999 10:25:33 +0000 (10:25 +0000)]
1. Run all pg_dump queries in single serializable transaction.
2. Get rid of locking when updating statistics in vacuum.
3. Use QuerySnapshot in COPY TO and call SetQuerySnashot
in main tcop loop before FETCH and COPY TO.
Tom Lane [Sat, 29 May 1999 01:48:06 +0000 (01:48 +0000)]
new_relation_targetlist used to cause about 8 separate (and
redundant) SearchSysCache searches per table column in an INSERT, which
accounted for a good percentage of the CPU time for INSERT ... VALUES().
Now it only does two searches in the typical case.
Tom Lane [Fri, 28 May 1999 17:03:31 +0000 (17:03 +0000)]
Repair performance problem in SI segment manipulations: iterating
through MAXBACKENDS array entries used to be fine when MAXBACKENDS = 64.
It's not so cool with MAXBACKENDS = 1024 (or more!), especially not in a
frequently-used routine like SIDelExpiredDataEntries. Repair by making
procState array size be the soft MaxBackends limit rather than the hard
limit, and by converting SIGetProcStateLimit() to a macro.
Tom Lane [Fri, 28 May 1999 01:54:53 +0000 (01:54 +0000)]
When closure of the backend connection is detected during pqFlush,
do the right thing: look for a NOTICE message from the backend before we
close our side of the socket. 6.4 libpq did not reliably print the backend's
hara-kiri message, 'The Postmaster has informed me ...', because it only
did the right thing if connection closure was detected during a read
attempt instead of a write attempt.
Tom Lane [Wed, 26 May 1999 22:57:39 +0000 (22:57 +0000)]
Patch from Andreas: when CREATE TABLE is followed by CREATE INDEX
before any tuples are loaded, preserve the default '1000 tuples' table
size estimate.
Tom Lane [Wed, 26 May 1999 21:51:13 +0000 (21:51 +0000)]
Fix pg_dump to use the same maximum-query-size constant as
the backend does. Remove unnecessary limitation on field size in
dumpClasses_dumpData (ie, -d or -D case).
Add chapters on CVS access, MVCC, SQL theory to the docs.
Add an appendix with more details on date/time attributes and handling.
Update most references to Postgres version numbers to 6.5,
*except* for the porting list which will require a report
from a successful installation to be updated.
Vadim B. Mikheev [Tue, 25 May 1999 18:31:28 +0000 (18:31 +0000)]
Get rid of page-level locking in btree-s.
BT_READ/BT_WRITE are BUFFER_LOCK_SHARE/BUFFER_LOCK_EXCLUSIVE now.
Also get rid of #define BT_VERSION_1 - we use version 1 as default
for near two years now.
Vadim B. Mikheev [Tue, 25 May 1999 18:20:31 +0000 (18:20 +0000)]
Get rid of page-level locking in btree-s.
LockBuffer is used to acquire read/write access
to index pages. Pages are released before leaving
index internals.
Jan Wieck [Tue, 25 May 1999 13:16:10 +0000 (13:16 +0000)]
Bugfix - Range table entries that are unused after rewriting should
not be marked inFromCl any longer. Otherwise the planner gets confused
and joins over them where in fact it does not have to.
Adjust hasSubLinks now with a recursive lookup - could be wrong in
multi action rules because parse state isn't reset correctly and all
actions in the rule are marked hasSubLinks if one of them has.