Etsuro Fujita

ExecOnConflictUpdate() passed t_ctid of the to-be-updated tuple to
ExecUpdate(). That's problematic primarily because of two reason: First
and foremost t_ctid could point to a different tuple. Secondly, and
that's what triggered the complaint by Stanislav, t_ctid is changed by
heap_update() to point to the new tuple version.  The behavior of AFTER
UPDATE triggers was therefore broken, with NEW.* and OLD.* tuples
spuriously identical within AFTER UPDATE triggers.

To fix both issues, pass a pointer to t_self of a on-stack HeapTuple
instead.

Fixing this bug lead to one change in regression tests, which previously
failed due to the first issue mentioned above. There's a reasonable
expectation that test fails, as it updates one row repeatedly within one
INSERT ... ON CONFLICT statement. That is only possible if the second
update is triggered via ON CONFLICT ... SET, ON CONFLICT ... WHERE, or
by a WITH CHECK expression, as those are executed after
ExecOnConflictUpdate() does a visibility check. That could easily be
prohibited, but given it's allowed for plain UPDATEs and a rare corner
case, it doesn't seem worthwhile.

Reported-By: Stanislav Grozev
Author: Andres Freund and Peter Geoghegan
Discussion: CAA78GVqy1+LisN-8DygekD_Ldfy=BJLarSpjGhytOsgkpMavfQ@mail.gmail.com
Backpatch: 9.5, where ON CONFLICT was introduced

At the end of crash recovery, unlogged relations are reset to the empty
state, using their init fork as the template. The init fork is copied to
the main fork without going through shared buffers. Unfortunately WAL
replay so far has not necessarily flushed writes from shared buffers to
disk at that point. In normal crash recovery, and before the
introduction of 'fast promotions' in fd4ced52 / 9.3, the
END_OF_RECOVERY checkpoint flushes the buffers out in time. But with
fast promotions that's not the case anymore.

To fix, force WAL writes targeting the init fork to be flushed
immediately (using the new FlushOneBuffer() function). In 9.5+ that
flush can centrally be triggered from the code dealing with restoring
full page writes (XLogReadBufferForRedoExtended), in earlier releases
that responsibility is in the hands of XLOG_HEAP_NEWPAGE's replay
function.

Backpatch to 9.1, even if this currently is only known to trigger in
9.3+. Flushing earlier is more robust, and it is advantageous to keep
the branches similar.

Typical symptoms of this bug are errors like
'ERROR:  index "..." contains unexpected zero page at block 0'
shortly after promoting a node.

Reported-By: Thom Brown
Author: Andres Freund and Michael Paquier
Discussion: 20150326175024.GJ451@alap3.anarazel.de
Backpatch: 9.1-

Commit 32f15d05 fixed this in configure, but missed the similar check
in the MSVC scripts.

Michael Paquier, per report from Victor Wagner

Peter Geoghegan

The original parallel sequential scan commit included only very limited
changes to the EXPLAIN output.  Aggregated totals from all workers were
displayed, but there was no way to see what each individual worker did
or to distinguish the effort made by the workers from the effort made by
the leader.

Per a gripe by Thom Brown (and maybe others).  Patch by me, reviewed
by Amit Kapila.

The single linked list of memory contexts could result in O(N^2)
performance to free a set of contexts if they were not freed in
reverse order of creation.  In many cases the reverse order was
used, but there were some significant exceptions that caused real-
world performance problems.  Rather than requiring all callers to
care about the order in which contexts were freed, and hunting down
and changing all existing cases where the wrong order was used, we
add one pointer per memory context so that the implementation
details are not so visible.

Jan Wieck

Commit 344cdff2 made failure to open the target of --output fatal.
For consistency, the --log-file switch should behave similarly.
Like the previous commit, back-patch to 9.5 but no further.

Daniel Verite

For unclear reasons, this function doesn't always read the expected data
in some old Perl versions.  Rewriting it to avoid use of ARGV seems to
dodge the problem, and this version is clearer anyway if you ask me.

In passing, also improve error message in adjacent append_to_file function.

To support this, we must reconcile some historical anomalies in the
behavior of -c.  In particular, as a backward-incompatibility, -c no
longer implies --no-psqlrc.

Pavel Stehule (code) and Catalin Iacob (documentation).  Review by
Michael Paquier and myself.  Proposed behavior per Tom Lane.

Commit e7cb7ee1 provided basic
infrastructure for allowing a foreign data wrapper or custom scan
provider to replace a join of one or more tables with a scan.
However, this infrastructure failed to take into account the need
for possible EvalPlanQual rechecks, and ExecScanFetch would fail
an assertion (or just overwrite memory) if such a check was attempted
for a plan containing a pushed-down join.  To fix, adjust the EPQ
machinery to skip some processing steps when scanrelid == 0, making
those the responsibility of scan's recheck method, which also has
the responsibility in this case of correctly populating the relevant
slot.

To allow foreign scans to gain control in the right place to make
use of this new facility, add a new, optional RecheckForeignScan
method.  Also, allow a foreign scan to have a child plan, which can
be used to correctly populate the slot (or perhaps for something
else, but this is the only use currently envisioned).

KaiGai Kohei, reviewed by Robert Haas, Etsuro Fujita, and Kyotaro
Horiguchi.

While convincing myself that commit 7e19db0c would solve both of
the problems recently reported by Andreas Seltenreich, I realized that
add_paths_to_joinrel's handling of LATERAL restrictions could be made
noticeably simpler and faster if we were to retain the minimum possible
parameterization for each joinrel (that is, the set of relids supplying
unsatisfied lateral references in it).  We already retain that for
baserels, in RelOptInfo.lateral_relids, so we can use that field for
joinrels too.

I re-pgindent'd the files touched here, which affects some unrelated
comments.

This is, I believe, just a minor optimization not a bug fix, so no
back-patch.

Per note from Tom Lane

It was possible for the planner to decide to join a LATERAL subquery to
the outer side of an outer join before the outer join itself is completed.
Normally that's fine because of the associativity rules, but it doesn't
work if the subquery contains a lateral reference to the inner side of the
outer join.  In such a situation the outer join *must* be done first.
join_is_legal() missed this consideration and would allow the join to be
attempted, but the actual path-building code correctly decided that no
valid join path could be made, sometimes leading to planner errors such as
"failed to build any N-way joins".

Per report from Andreas Seltenreich.  Back-patch to 9.3 where LATERAL
support was added.

Noted by Tom Lane:
- PostgresNode had a BEGIN block which created files, contrary to
  perlmod suggestions to do that only on INIT blocks.
- Assign ports randomly rather than starting from 90600.

Noted by Noah Misch:
- Change use of no-longer-set PGPORT environment variable to $node->port
- Don't start a server in pg_controldata test
- PostgresNode was reading the PID file incorrectly; test the right
  thing, and chomp the line we read from the PID file.
- Remove an unused $devnull variable
- Use 'pg_ctl kill' instead of "kill" directly, for Windos portability.
- Make server log names more informative.

Author: Michael Paquier

Commit d04c8ed9 added another support function to the GIST API,
but overlooked mentioning it in xindex.sgml's summary of index support
functions.

Anastasia Lubennikova

This way, existing .gitignore entries and makefile clean actions will
automatically apply to the tempdir, should it survive a TAP test run
(which can happen if the user control-C's out of the run, for example).

Michael Paquier, per a complaint from me

This should make Coverity deduce that plperl_call_perl_func() does not
dereference NULL argtypes.  Back-patch to 9.5, where the affected code
was introduced.

Michael Paquier

In commit 1ea0c73c I added a section to user-manag.sgml about how to drop
roles that own objects; but as pointed out by Stephen Frost, I neglected
that shared objects (databases or tablespaces) may need special treatment.
Fix that.  Back-patch to supported versions, like the previous patch.

As pointed out by Fujii Masao, we weren't quite there on a standby
behaving sanely: first because we were failing to acquire the correct
state in the case where no XLOG_PARAMETER_CHANGE message was sent
(because a checkpoint had already happened after the setting was changed
in the master, and then the standby was restarted); and second because
promoting the standby with the feature enabled failed to activate it if
the master had the feature disabled.

This patch fixes both those misbehaviors hopefully without
re-introducing any old problems.

Also change the hint emitted in a standby together with the error
message about the feature being disabled, to make it point out that the
place to chance the setting is the master.  Otherwise, if the setting is
already enabled in the standby, it is very confusing to have it say that
the setting must be enabled ...

Authors: Álvaro Herrera, Petr Jelínek.
Backpatch to 9.5.

Formerly, if "psql -o foo" failed to open the output file "foo", it would
print an error message but then carry on as though -o had not been
specified at all.  This seems contrary to expectation: a program that
cannot open its output file normally fails altogether.  Make psql do
exit(1) after reporting the error.

If "\o foo" failed to open "foo", it would print an error message but then
reset the output file to stdout, as if the argument had been omitted.
This is likewise pretty surprising behavior.  Make it keep the previous
output state, instead.

psql keeps SIGPIPE interrupts disabled when it is writing to a pipe, either
a pipe specified by -o/\o or a transient pipe opened for purposes such as
using a pager on query output.  The logic for this was too simple and could
sometimes re-enable SIGPIPE when a -o pipe was still active, thus possibly
leading to an unexpected psql crash later.

Fixing the last point required getting rid of the kluge in PrintQueryTuples
and ExecQueryUsingCursor whereby they'd transiently change the global
queryFout state, but that seems like good cleanup anyway.

Back-patch to 9.5 but not further; these are minor-enough issues that
changing the behavior in stable branches doesn't seem appropriate.

Michael Paquier

The formatting modes that depend on knowledge of the terminal window width
did not work right when printing a query result that's been fetched in
sections (as a result of FETCH_SIZE).  ExecQueryUsingCursor() would force
use of the pager as soon as there's more than one result section, and then
print.c would see an output file pointer that's not stdout and incorrectly
conclude that the terminal window width isn't relevant.

This has been broken all along for non-expanded "wrapped" output format,
and as of 9.5 the issue affects expanded mode as well.  The problem also
caused "\pset expanded auto" mode to invariably *not* switch to expanded
output in a segmented result, which seems to me to be exactly backwards.

To fix, we need to pass down an "is_pager" flag to inform the print.c
subroutines that some calling level has already replaced stdout with a
pager pipe, so they should (a) not do that again and (b) nonetheless honor
the window size.  (Notably, this makes the first is_pager test in
print_aligned_text() not be dead code anymore.)

This patch is a bit invasive because there are so many existing calls of
printQuery()/printTable(), but fortunately all but a couple can just pass
"false" for the added parameter.

Back-patch to 9.5 but no further.  Given the lack of field complaints,
it's not clear that we should change the behavior in stable branches.
Also, the API change for printQuery()/printTable() might possibly break
third-party code, again something we don't like to do in stable branches.
However, it's not quite too late to do this in 9.5, and with the larger
scope of the problem there, it seems worth doing.

The original code was a bit clunky; make it more amenable for further
reuse by creating a new Perl package PostgresNode, which is an
object-oriented representation of a single server, with some support
routines such as init, start, stop, psql.  This serves as a better basis
on which to build further test code, and enables writing tests that use
more than one server without too much complication.

This commit modifies a lot of the existing test files, mostly to remove
explicit calls to system commands (pg_ctl) replacing them with method
calls of a PostgresNode object.  The result is quite a bit more
straightforward.

Also move some initialization code to BEGIN and INIT blocks instead of
having it straight in as top-level code.

This commit also introduces package RecursiveCopy so that we can copy
whole directories without having to depend on packages that may not be
present on vanilla Perl 5.8 installations.

I also ran perltidy on the modified files, which changes some code sites
that are not otherwise touched by this patch.  I tried to avoid this,
but it ended up being more trouble than it's worth.

Authors: Michael Paquier, Álvaro Herrera
Review: Noah Misch

Peter Geoghegan

We tried to fetch statistics data from the index metapage, which does not
work if the index isn't actually present.  If the index is hypothetical,
instead extrapolate some plausible internal statistics based on the index
page count provided by the index-advisor plugin.

There was already some code in gincostestimate() to invent internal stats
in this way, but since it was only meant as a stopgap for pre-9.1 GIN
indexes that hadn't been vacuumed since upgrading, it was pretty crude.
If we want it to support index advisors, we should try a little harder.
A small amount of testing says that it's better to estimate the entry pages
as 90% of the index, not 100%.  Also, estimating the number of entries
(keys) as equal to the heap tuple count could be wildly wrong in either
direction.  Instead, let's estimate 100 entries per entry page.

Perhaps someday somebody will want the index advisor to be able to provide
these numbers more directly, but for the moment this should serve.

Problem report and initial patch by Julien Rouhaud; modified by me to
invent less-bogus internal statistics.  Back-patch to all supported
branches, since we've supported index advisors since 9.0.

Don't force the data width to extend all the way to the right margin if it
doesn't need to.  This reverts the behavior in non-wrapping cases to be
what it was in 9.4.  Also, make the logic that ensures the data line width
is at least equal to the record-header line width a little less obscure.

In passing, avoid possible calculation of log10(0).  Probably that's
harmless, given the lack of field complaints, but it seems risky:
conversion of NaN to an integer isn't well defined.

The previous coding could overrun the provided buffer size for a very large
input, or lose precision for a very small input.  Adopt the methodology
that's been in use in the equivalent backend code for a long time.

Per private report from Bas van Schaik.  Back-patch to all supported
branches.

We should ignore output_columns unless it's greater than zero.
A zero means we couldn't get any information from ioctl(TIOCGWINSZ);
in that case the expected behavior is to print the data at native width,
not to wrap it at the smallest possible value.  print_aligned_text()
gets this consideration right, but print_aligned_vertical() lost track
of this detail somewhere along the line.

Allow pg_rewind to work when target timeline was switched. Now
user can return promoted standby to old master.

Target timeline history becomes a global variable. Index
in target timeline history is used in function interfaces instead of
specifying TLI directly. Thus, SimpleXLogPageRead() can easily start
reading XLOGs from next timeline when current timeline ends.

Author: Alexander Korotkov
Review: Michael Paquier

This area was rather heavily whacked around in 6513633b and follow-on
commits, and it was showing it, because the logic to calculate the
allowable data width in wrapped expanded mode had only the vaguest
relationship to the logic that was actually printing the data.  It was
not very close to being right about the conditions requiring overhead
columns to be added.  Aside from being wrong, it was pretty unreadable
and under-commented.  Rewrite it so it corresponds to what the printing
code actually does.

In passing, remove a couple of dead tests in the printing logic, too.

Per a complaint from Jeff Janes, though this doesn't look much like his
patch because it fixes a number of other corner-case bogosities too.
One such fix that's visible in the regression test results is that
although the code was attempting to enforce a minimum data width of
3 columns, it sometimes left less space than that available.

It's amazing how fast things become obsolete these days.

Amit Langote

In commit 8abb3cda I attempted to cache
the expression state trees constructed for domain CHECK constraints for
the life of the backend (assuming the domain's constraints don't get
redefined).  However, this turns out not to work very well, because
execQual.c will run those state trees with ecxt_per_query_memory pointing
to a query-lifespan context, and in some situations we'll end up with
pointers into that context getting stored into the state trees.  This
happens in particular with SQL-language functions, as reported by
Emre Hasegeli, but there are many other cases.

To fix, keep only the expression plan trees for domain CHECK constraints
in the typcache's data structure, and revert to performing ExecInitExpr
(at least) once per query to set up expression state trees in the query's
context.

Eventually it'd be nice to undo this, but that will require some careful
thought about memory management for expression state trees, and it seems
far too late for any such redesign in 9.5.  This way is still much more
efficient than what happened before 8abb3cda.

Previously, we did many conversions for Cyrillic and Central European
single-byte encodings by converting to a related MULE_INTERNAL coding
scheme before converting to the destination.  This seems unnecessarily
inefficient.  Moreover, if the conversion encounters an untranslatable
character, the error message will confusingly complain about failure
to convert to or from MULE_INTERNAL, rather than the user-visible
encodings.  Worse still, this approach results in some completely
unnecessary conversion failures; there are cases where the chosen
MULE subset lacks characters that exist in both of the user-visible
encodings, causing a conversion failure that need not occur.

This patch fixes the first two of those deficiencies by introducing
a new local2local() conversion support subroutine for direct conversion
between any two single-byte character sets, and adding new conversion
tables where needed.  However, I generated the new conversion tables by
testing PG 9.5's behavior, so that the actual conversion behavior is
bug-compatible with previous releases; the only user-visible behavior
change is that the error messages for conversion failures are saner.
Changes in the conversion behavior will probably ensue after discussion.

Interestingly, although this approach requires more tables, the .so files
actually end up smaller (at least on my x86_64 machine); the tables are
smaller than the management code needed for double conversion.

Per a complaint from Albe Laurenz.

Some of the Unicode/*.map files had identification comments added to them,
evidently by hand.  Others did not.  Modify the generating scripts to
produce these comments automatically, and update the generated files that
lacked them.

This is just minor cleanup as a by-product of trying to verify that the
*.map files can indeed be reproduced from authoritative data.  There are a
depressingly large number that fail to reproduce from the claimed sources.
I have not touched those in this commit, except for the JIS 2004-related
files which required only a single comment update to match.

Since this only affects comments, no need to consider a back-patch.

Previously, if no host information had been specified at connection time,
PQhost() would return NULL (unless you are on Windows, in which case you
got "localhost").  This is an unhelpful definition for a couple of reasons:
it can cause corner-case crashes in applications (cf commit c5ef8ce5),
and there's no well-defined way for applications to find out the socket
directory path that's actually in use.  As an example of the latter
problem, psql substituted DEFAULT_PGSOCKET_DIR for NULL in a couple of
places, but this is subtly wrong because it's conceivable that psql is
using a libpq shared library that was built with a different setting.

Hence, change PQhost() to return DEFAULT_PGSOCKET_DIR when appropriate,
and strip out the now-dead substitutions in psql.  (There is still one
remaining reference to DEFAULT_PGSOCKET_DIR in psql, in prompt.c, which
I don't see a nice way to get rid of.  But it only controls a prompt
abbreviation decision, so it seems noncritical.)

Also update the docs for PQhost, which had never previously mentioned
the possibility of a socket directory path being returned.  In passing
fix the outright-incorrect code comment about PGconn.pgunixsocket.