Skip to content

G
Gpdb

Greenplum / Gpdb

通知 7
Star 1
Fork 0
G
Gpdb

体验新版 GitCode，发现更多精彩内容 >>

切换分支/标签
  1. 30 6月, 2008 1 次提交
  3. 29 6月, 2008 1 次提交
  5. 28 6月, 2008 2 次提交
    • T
      Consider a clause to be outerjoin_delayed if it references the nullable side · dcc23347
      由 Tom Lane 提交于 
      of any lower outer join, even if it also references the non-nullable side and
so could not get pushed below the outer join anyway.  We need this in case
the clause is an OR clause: if it doesn't get marked outerjoin_delayed,
create_or_index_quals() could pull an indexable restriction for the nullable
side out of it, leading to wrong results as demonstrated by today's bug
report from toruvinn.  (See added regression test case for an example.)

In principle this has been wrong for quite a while.  In practice I don't
think any branch before 8.3 can really show the failure, because
create_or_index_quals() will only pull out indexable conditions, and before
8.3 those were always strict.  So though we might have improperly generated
null-extended rows in the outer join, they'd get discarded from the result
anyway.  The gating factor that makes the failure visible is that 8.3
considers "col IS NULL" to be indexable.  Hence I'm not going to risk
back-patching further than 8.3.
      dcc23347
    • M
      Fix standalone libpq build on win32. · f6c1dece
      由 Magnus Hagander 提交于 
      Hiroshi Saito
      f6c1dece
  7. 27 6月, 2008 9 次提交
  9. 26 6月, 2008 9 次提交
  11. 25 6月, 2008 1 次提交
    • T
      Reduce the alignment requirement of type "name" from int to char, and arrange · 5f6f840e
      由 Tom Lane 提交于 
      to suppress zero-padding of "name" entries in indexes.

The alignment change is unlikely to save any space, but it is really needed
anyway to make the world safe for our widespread practice of passing plain
old C strings to functions that are declared as taking Name.  In the previous
coding, the C compiler was entitled to assume that a Name pointer was
word-aligned; but we were failing to guarantee that.  I think the reason
we'd not seen failures is that usually the only thing that gets done with
such a pointer is strcmp(), which is hard to optimize in a way that exploits
word-alignment.  Still, some enterprising compiler guy will probably think
of a way eventually, or we might change our code in a way that exposes
more-obvious optimization opportunities.

The padding change is accomplished in one-liner fashion by declaring the
"name" index opclasses to use storage type "cstring" in pg_opclass.h.
Normally btree and hash don't allow a nondefault storage type, because they
don't have any provisions for converting the input datum to another type.
However, because name and cstring are effectively the same thing except for
padding, no conversion is needed --- we only need index_form_tuple() to treat
the datum as being cstring not name, and this is sufficient.  This seems to
make for about a one-third reduction in the typical sizes of system catalog
indexes that involve "name" columns, of which we have many.

These two changes are only weakly related, but the alignment change makes
me feel safer that the padding change won't introduce problems, so I'm
committing them together.
      5f6f840e
  13. 24 6月, 2008 11 次提交
  15. 20 6月, 2008 3 次提交
    • T
      Seems I was too optimistic in supposing that sinval's maxMsgNum could be · dab421d2
      由 Tom Lane 提交于 
      read and written without a lock.  The value itself is atomic, sure, but on
processors with weak memory ordering it's possible for a reader to see the
value change before it sees the associated message written into the buffer
array.  Fix by introducing a spinlock that's used just to read and write
maxMsgNum.  (We could do this with less overhead if we recognized a concept
of "memory access barrier"; is it worth introducing such a thing?  At the
moment probably not --- I can't measure any clear slowdown from adding the
spinlock, so this solution is probably fine.)  Per buildfarm results.
      dab421d2
    • T
      Rewrite the sinval messaging mechanism to reduce contention and avoid · fad153ec
      由 Tom Lane 提交于 
      unnecessary cache resets.  The major changes are:

* When the queue overflows, we only issue a cache reset to the specific
backend or backends that still haven't read the oldest message, rather
than resetting everyone as in the original coding.

* When we observe backend(s) falling well behind, we signal SIGUSR1
to only one backend, the one that is furthest behind and doesn't already
have a signal outstanding for it.  When it finishes catching up, it will
in turn signal SIGUSR1 to the next-furthest-back guy, if there is one that
is far enough behind to justify a signal.  The PMSIGNAL_WAKEN_CHILDREN
mechanism is removed.

* We don't attempt to clean out dead messages after every message-receipt
operation; rather, we do it on the insertion side, and only when the queue
fullness passes certain thresholds.

* Split SInvalLock into SInvalReadLock and SInvalWriteLock so that readers
don't block writers nor vice versa (except during the infrequent queue
cleanout operations).

* Transfer multiple sinval messages for each acquisition of a read or
write lock.
      fad153ec
    • T
      Fix a few places that were non-multibyte-safe in tsearch configuration file · 30dc388a
      由 Tom Lane 提交于 
      parsing.  Per bug #4253 from Giorgio Valoti.
      30dc388a
  17. 19 6月, 2008 3 次提交