| Commit message (Collapse) | Author | Age |
|---|
| |
|
|
|
|
|
|
|
|
| |
so that we close and flush the doomed relation's relcache entry before
we start to delete the underlying catalog rows, rather than afterwards.
For awhile yesterday I thought that an unexpected relcache entry rebuild
partway through this sequence might explain the infrequent parallel
regression failures we were chasing. It doesn't, mainly because there's
no CommandCounterIncrement in the sequence and so the deletions aren't
"really" done yet. But it sure seems like trouble waiting to happen.
|
| |
|
|
| |
Magnus Hagander
|
| |
|
|
|
|
|
|
|
| |
relcache entries. Also, change TransactionIdIsCurrentTransactionId()
so that if consulted during transaction abort, it will not say that
the aborted xact is still current. (It would be better to ensure that
it's never called at all during abort, but I'm not sure we can easily
guarantee that.) In combination, these fix a crash we have seen
occasionally during parallel regression tests of 8.0.
|
| |
|
|
|
|
|
|
| |
from being accepted after the outer right brace. Per report from
Markus Bertheau.
Also add regression test cases for this change, and for previous
recent array literal parser changes.
|
| |
|
|
| |
module isn't fully initialized yet.
|
| |
|
|
|
| |
if the target PID is a PG postmaster or backend --- for our purposes that
is actually better than the Unix behavior. Per Dave Page and Andrew Dunstan.
|
| |
|
|
|
|
|
|
|
| |
PROCLOCK structs in shared memory now have only a bitmask for held
locks, rather than counts (making them 40 bytes smaller, which is a
good thing). Multiple locks within a transaction are counted in the
local hash table instead, and we have provision for tracking which
ResourceOwner each count belongs to. Solves recently reported problem
with memory leakage within long transactions.
|
| |
|
|
| |
changes.
|
| |
|
|
| |
CIDR column in examples first.
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
for every command executed within a transaction. For long transactions
this was a significant memory leak. Instead, we can delete a portal's
or subtransaction's ResourceOwner immediately, if we physically transfer
the information about its locks up to the parent owner. This does not
fully solve the leak problem; we need to do something about counting
multiple acquisitions of the same lock in order to fix it. But it's a
necessary step along the way.
|
| | |
|
| |
|
|
|
|
| |
ColLabel instead of just ColId --- that is, any keyword can appear after
a dot and it will be taken as an identifier. Fixes problems with names
that are okay as standalone function names but fail when qualified.
|
| |
|
|
|
|
|
| |
updates are no longer WAL-logged nor even fsync'd; we do not need to,
since after a crash no old pg_subtrans data is needed again. We truncate
pg_subtrans to RecentGlobalXmin at each checkpoint. slru.c's API is
refactored a little bit to separate out the necessary decisions.
|
| |
|
|
|
|
|
|
|
|
| |
RecentXmin (== MyProc->xmin). This ensures that it will be safe to
truncate pg_subtrans at RecentGlobalXmin, which should largely eliminate
any fear of bloat. Along the way, eliminate SubTransXidsHaveCommonAncestor,
which isn't really needed and could not give a trustworthy result anyway
under the lookback restriction.
In an unrelated but nearby change, #ifdef out GetUndoRecPtr, which has
been dead code since 2001 and seems unlikely to ever be resurrected.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
>>'127.0.0.1/32' instead of '127.0.0.1 255.255.255.255'.
>>
>>
>
>Yeah, that's probably the path of least resistance. Note that the
>comments and possibly the SGML docs need to be adjusted to match,
>however, so it's not quite a one-liner.
Andrew Dunstan
|
| |
|
|
| |
Gavin Sherry
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
> why does CVS tip still give me
>
> regression=# select extract(century from now());
> date_part
> -----------
> 20
> (1 row)
> [ ... looks in code ... ]
>
> Apparently it's because you fixed only timestamp_part, and not
> timestamptz_part. I'm not too sure about what timestamp_trunc or
> timestamptz_trunc should do, but they may be wrong as well.
Sigh... as usual, what is not tested does not work:-(
> Could we have a more complete patch?
Please find a submission attached. I hope it really fixes all decade,
century and millenium issues for extract and *_trunc functions on
interval
and other timestamp types. If someone could check that the results
are reasonnable, it would be great.
I indeed overlooked the fact that there were two functions. The patch
fixes the code so that both variants agree.
I added comments to interval extractions, because it relies on the C
division to have a negative remainder: -7/10 = 0 and remains -7.
As for *_trunc functions, I have chosen to put the first year of the
century or millennium: -100, 1, 101... 1001 2001 etc. Indeed, I don't
think it would make sense to put 2000 (last year of the 2nd millennium)
for rounding all years of the third millenium.
I also fixed the code so that all decades last 10 years and decade 199
means the 1990's.
I have added some tests that are relevant to deal with tricky cases. The
formula may be simplified, but all these cases must pass. Please keep
them.
Fabien Coelho
|
| |
|
|
|
|
|
|
|
|
|
|
| |
presence of dropped columns. Document the already-presumed fact that
eref aliases in relation RTEs are supposed to have entries for dropped
columns; cause the user alias structs to have such entries too, so that
there's always a one-to-one mapping to the underlying physical attnums.
Adjust expandRTE() and related code to handle the case where a column
that is part of a JOIN has been dropped. Generalize expandRTE()'s API
so that it can be used in a couple of places that formerly rolled their
own implementation of the same logic. Fix ruleutils.c to suppress
display of aliases for columns that were dropped since the rule was made.
|
| |
|
|
|
|
|
|
|
|
| |
value of 'start' could be past the end of the page, if the page was
split by some concurrent inserting process since we visited it. In
this situation the code could look at bogus entries and possibly find
a match (since after all those entries still contain what they had
before the split). This would lead to 'specified item offset is too large'
followed by 'PANIC: failed to add item to the page', as reported by Joe
Conway for scenarios involving heavy concurrent insertion activity.
|
| |
|
|
|
|
|
| |
to the physical layout of the rowtype, ie, there are dummy arguments
corresponding to any dropped columns in the rowtype. We formerly had a
couple of places that did it this way and several others that did not.
Fixes Gaetano Mendola's "cache lookup failed for type 0" bug of 5-Aug.
|
| |
|
|
| |
proposals by Oliver Jowett. Update documentation.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
of XLogInsert had the same sort of checkpoint interlock problem as
RecordTransactionCommit, and indeed I found some. Btree index build
and ALTER TABLE SET TABLESPACE write data outside the friendly confines
of the buffer manager, and therefore they have to take their own
responsibility for checkpoint interlock. The easiest solution seems to
be to force smgrimmedsync at the end of the index build or table copy,
even when the operation is being WAL-logged. This is sufficient since
the new index or table will be of interest to no one if we don't get
as far as committing the current transaction.
|
| |
|
|
|
|
| |
therefore starting with GetCurrentTransactionId is wrong. Fixes
miscomputation of RecentGlobalXmin leading to bizarre behavior
reported by Gavin Sherry.
|
| |
|
|
| |
Oliver Jowett.
|
| | |
|
| | |
|
| |
|
|
| |
patches.
|
| |
|
|
|
|
| |
don't hold an open file reference to the original table at the end.
This is a good thing in any case, particularly so on Windows which
cannot drop the table file otherwise.
|
| |
|
|
|
|
| |
by the SQL standard. For backwards compatibility, however, continue to
accept the syntax without. Minor editorialization in the reference pages
for these commands, too.
|
| |
|
|
|
|
| |
compliance with SQL2003 spec syntax.
Oliver Jowett
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
and doesn't process forward slashes in the same way as external
commands. Quoting the first argument to COPY does not convert forward
to backward slashes, but COPY does properly process quoted forward
slashes in the second argument.
Win32 COPY works with quoted forward slashes in the first argument only if the
current directory is the same as the directory of the first argument.
|
| |
|
|
| |
that refers to a now-deleted userid. Per gripe from Chris Ochs.
|
| |
|
|
| |
Thanks to Michael Fuhr for identifying the problem.
|
| | |
|
| |
|
|
|
|
| |
for transaction commits that occurred just before the checkpoint. This is
an EXTREMELY serious bug --- kudos to Satoshi Okada for creating a
reproducible test case to prove its existence.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
slashes to backslashes #ifdef WIN32. This is to cope with the fact
that Windows seems exceedingly unfriendly to slashes in shell commands,
as per recent discussion.
|
| |
|
|
| |
list. More consistent with our other list-containing GUC variables.
|
| |
|
|
| |
on other recent changes in runtime parameter list.
|
| | |
|
| |
|
|
|
|
|
| |
CurrentMemoryContext is DLLIMPORT on Win32. Work around that by
creating stubs in the backend for palloc/pstrdup.
Also fix pg_dumpall to do proper quoting on Win32.
|
| |
|
|
|
|
|
|
|
| |
was previously allowed in odd places with odd results now causes an ERROR.
Also changed behavior with respect to whitespace -- trailing whitespace is
now ignored as well as leading whitespace (which has always been ignored).
Documentation updated to reflect change in whitespace handling. Also some
refactoring to what I believe is a more sensible order of several paragraphs.
|
| | |
|
