Tom Lane [Mon, 22 Nov 1999 02:06:31 +0000 (02:06 +0000)]
Clean up some problems in error recovery --- elog() was pretty broken
for the case of errors in backend startup, and proc_exit's method for
coping with errors during proc_exit was *completely* busted. Fixed per
discussions on pghackers around 11/6/99.
Tom Lane [Mon, 22 Nov 1999 02:03:21 +0000 (02:03 +0000)]
Come to think of it, DropBuffers() could have the same problem as
ReleaseRelationBuffers --- need to wait if anyone is trying to flush
out that buffer.
Tom Lane [Mon, 22 Nov 1999 01:19:42 +0000 (01:19 +0000)]
ReleaseRelationBuffers() failed to check for I/O in progress on a buffer
it wants to release. This leads to a race condition: does the backend
that's trying to flush the buffer do so before the one that's deleting the
relation does so? Usually no problem, I expect, but on occasion this could
lead to hard-to-reproduce complaints from md.c, especially mdblindwrt.
Tom Lane [Sun, 21 Nov 1999 23:25:47 +0000 (23:25 +0000)]
Combine index_info and find_secondary_indexes into a single routine that
returns a list of RelOptInfos, eliminating the need for static state
in index_info. That static state was a direct cause of coredumps; if
anything decided to elog(ERROR) partway through an index_info search of
pg_index, the next query would try to close a scan pointer that was
pointing at no-longer-valid memory. Another example of the reasons to
avoid static state variables...
Tom Lane [Sun, 21 Nov 1999 20:01:10 +0000 (20:01 +0000)]
index_destroy() must grab exclusive access to the parent table
of the index it wants to destroy. This ensures that no other backend is
actively scanning or updating that index. Getting exclusive access on
the index alone is NOT sufficient, because the executor is rather
cavalier about getting locks on indexes --- see ExecOpenIndices().
It might be better to grab index locks in the executor, but I'm not
sure the extra lockmanager traffic is really worth it just to make
index_destroy cleaner.
Tom Lane [Sun, 21 Nov 1999 04:16:17 +0000 (04:16 +0000)]
Change backend-side COPY to write files with permissions 644 not 666
(whoever thought world-writable files were a good default????). Modify
the pg_pwd code so that pg_pwd is created with 600 permissions. Modify
initdb so that permissions on a pre-existing PGDATA directory are not
blindly accepted: if the dir is already there, it does chmod go-rwx
to be sure that the permissions are OK and the dir actually is owned
by postgres.
Tom Lane [Sun, 21 Nov 1999 01:58:22 +0000 (01:58 +0000)]
Repair problem exposed by Jan's new parallel-regression-test scaffold:
inval.c thought it could safely use the catcache to look up the OIDs of
system relations. Not good, considering that inval.c could be called
during catcache loading, if a shared-inval message arrives. Rip out the
lookup logic and instead use the known OIDs from pg_class.h.
Tom Lane [Sun, 21 Nov 1999 01:53:39 +0000 (01:53 +0000)]
Tweak run_check.sh so it prints the name of each test in a parallel group
as that test finishes --- helps to give the impression that something is
happening...
Tom Lane [Sat, 20 Nov 1999 21:39:36 +0000 (21:39 +0000)]
Fix problems with CURRENT_DATE and related functions being used in
table defaults or rules: translate them to a function call so that
parse_coerce doesn't reduce them to a date or time constant immediately.
Also, eliminate a lot of redundancy in the expression grammar by
defining a new nonterminal com_expr, which contains all the productions
that can be shared by a_expr and b_expr.
Tom Lane [Tue, 16 Nov 1999 06:13:36 +0000 (06:13 +0000)]
Modify elog() logic so that it won't try to longjmp(Warn_restart) before
Warn_restart has been set by the backend main loop. This means that
elog(ERROR) or elog(FATAL) in the postmaster or during backend startup
now have well-defined behavior: proc_exit() rather than coredump.
In the case of elog() inside the postmaster, I think that proc_exit()
is probably not enough --- don't we want our child backends to be
forced to quit too? But I don't understand Vadim's recent changes in
this area, so I'll leave it to him to look over and tweak if needed.
Tom Lane [Mon, 15 Nov 1999 02:00:15 +0000 (02:00 +0000)]
Implement subselects in target lists. Also, relax requirement that
subselects can only appear on the righthand side of a binary operator.
That's still true for quantified predicates like x = ANY (SELECT ...),
but a subselect that delivers a single result can now appear anywhere
in an expression. This is implemented by changing EXPR_SUBLINK sublinks
to represent just the (SELECT ...) expression, without any 'left hand
side' or combining operator --- so they're now more like EXISTS_SUBLINK.
To handle the case of '(x, y, z) = (SELECT ...)', I added a new sublink
type MULTIEXPR_SUBLINK, which acts just like EXPR_SUBLINK used to.
But the grammar will only generate one for a multiple-left-hand-side
row expression.
Tom Lane [Fri, 12 Nov 1999 06:39:34 +0000 (06:39 +0000)]
Fix ExecSubPlan to handle nulls per the SQL spec --- it didn't combine
nulls with non-nulls using proper three-valued boolean logic. Also clean
up ExecQual to make it clearer that ExecQual *does* follow the SQL spec
for boolean nulls. See '[BUGS] (null) != (null)' thread around 10/26/99
for more detail.
Bruce Momjian [Thu, 11 Nov 1999 00:10:14 +0000 (00:10 +0000)]
In the spirit of TODO item
* Add use of 'const' for varibles in source tree
(which is misspelled, btw.)
I went through the front-end libpq code and did so. This affects in
particular the various accessor functions (such as PQdb() and
PQgetvalue()) as well as, by necessity, the internal helpers they use.
I have been really thorough in that regard, perhaps some people will find
it annoying that things like
char * foo = PQgetvalue(res, 0, 0)
will generate a warning. On the other hand it _should_ generate one. This
is no real compatibility break, although a few clients will have to be
fixed to suppress warnings. (Which again would be in the spirit of the
above TODO.)
In addition I replaced some int's by size_t's and removed some warnings
(and generated some new ones -- grmpf!). Also I rewrote PQoidStatus (so it
actually honors the const!) and supplied a new function PQoidValue that
returns a proper Oid type. This is only front-end stuff, none of the
communicaton stuff was touched.
The psql patch also adds some new consts to honor the new libpq situation,
as well as fixes a fatal condition that resulted when using the -V
(--version) option and there is no database listening.
So, to summarize, the psql you should definitely put in (with or without
the libpq). If you think I went too far with the const-mania in libpq, let
me know and I'll make adjustments. If you approve it, I will also update
the docs.
Tom Lane [Fri, 5 Nov 1999 06:43:45 +0000 (06:43 +0000)]
environment variable set by MULTIBYTE startup code should be
stored in malloc'd space, not in a static variable. Otherwise environment
variable list is corrupted if libpq is dynamically unlinked...
Hiroshi Inoue [Thu, 4 Nov 1999 08:01:09 +0000 (08:01 +0000)]
Make it possible to execute crashed CREATE/DROP commands again.
Now indexes of pg_class and pg_type are unique indexes
and guarantee the uniqueness of correponding attributes.
heap_create() was changed to take another boolean parameter
which allows to postpone the creation of disk file.
The name of rd_nonameunlinked was changed to rd_unlinked.
It is used generally(not only for noname relations) now.
Requires initdb.
Tom Lane [Mon, 1 Nov 1999 05:18:31 +0000 (05:18 +0000)]
Eliminate RewritePreprocessQuery, which was taking an
unreasonable amount of time to clean up after a vanished parser problem.
Don't call fireRIRonSubselect when we know there are no subselects,
either.
Tom Lane [Mon, 1 Nov 1999 05:15:13 +0000 (05:15 +0000)]
Eliminate some unbelievably cheesy code in _copyConst().
Apparently, back in the dim reaches of prehistory, the parser couldn't
be trusted to label Const nodes with the correct constbyval value ...
and someone preferred to patch around this in copyObject rather than
fix the problem at the source. The problem is long gone, but the hack
lingered on. Until now.
Tom Lane [Sat, 30 Oct 1999 23:13:30 +0000 (23:13 +0000)]
Avoid duplicate ExecTypeFromTL() call in ExecInitJunkFilter() by passing
in the TupleDesc that the caller already has (for call from ExecMain) or
can make just as easily as ExecInitJunkFilter() can (for call from
ExecAppend). Also, don't bother to build a junk filter for an INSERT
operation that doesn't actually need one, which is the normal case.
Tom Lane [Sat, 30 Oct 1999 23:07:55 +0000 (23:07 +0000)]
Skip invoking set_uppernode_references() for a RESULT node
that has no subplan --- saves a material amount of time for a simple
INSERT ... VALUES query.
Tom Lane [Sat, 30 Oct 1999 23:06:32 +0000 (23:06 +0000)]
Rewrite preprocess_targetlist() to reduce overhead for simple INSERTs.
In particular, don't bother to look up type information for attributes
where we're not actually going to use it, and avoid copying entire tlist
structure when it's not necessary.
Tom Lane [Sat, 30 Oct 1999 17:27:15 +0000 (17:27 +0000)]
Further performance improvements in sorting: reduce number of comparisons
during initial run formation by keeping both current run and next-run
tuples in the same heap (yup, Knuth is smarter than I am). And, during
merge passes, make use of available sort memory to load multiple tuples
from any one input 'tape' at a time, thereby improving locality of
access to the temp file.
Tom Lane [Sat, 30 Oct 1999 01:18:16 +0000 (01:18 +0000)]
Put back code in nodeAgg to generate a dummy all-nulls input tuple
before calling execProject, when the outerPlan has returned zero tuples.
I took this out under the mistaken impression that the input tuple
couldn't be referenced by execProject if we weren't in GROUP BY mode.
But it can, if we're in an UPDATE or DELETE...