Commit 894459e59f revealed this option to
be broken for NLS builds on Darwin, but "make -C contrib/unaccent check"
and the buildfarm client rely on it. Fix that configuration by
redefining the option to imply LANG=C on Darwin. In passing, use LANG=C
instead of LANG=en on Windows; since only postmaster startup uses that
value, testers are unlikely to notice the change. Back-patch to 9.0,
like the predecessor commit.
Up to now, the "child" executor state trees generated for EvalPlanQual
rechecks have simply shared the ResultRelInfo arrays used for the original
execution tree. However, this leads to dangling-pointer problems, because
ExecInitModifyTable() is all too willing to scribble on some fields of the
ResultRelInfo(s) even when it's being run in one of those child trees.
This trashes those fields from the perspective of the parent tree, because
even if the generated subtree is logically identical to what was in use in
the parent, it's in a memory context that will go away when we're done
with the child state tree.
We do however want to share information in the direction from the parent
down to the children; in particular, fields such as es_instrument *must*
be shared or we'll lose the stats arising from execution of the children.
So the simplest fix is to make a copy of the parent's ResultRelInfo array,
but not copy any fields back at end of child execution.
Per report from Manuel Kniep. The added isolation test is based on his
example. In an unpatched memory-clobber-enabled build it will reliably
fail with "ctid is NULL" errors in all branches back to 9.1, as a
consequence of junkfilter->jf_junkAttNo being overwritten with $7f7f.
This test cannot be run as-is before that for lack of WITH syntax; but
I have no doubt that some variant of this problem can arise in older
branches, so apply the code change all the way back.
The flag is supposed to be copied from the record. Same issue with
track_commit_timestamps, but that's master-only.
Report and fix by Petr Jalinek. Backpatch to 9.4, where wal_log_hints was
added.
The folly of the previous arrangement was just demonstrated: there's no
convenient way to add fields to ExplainState without breaking ABI, even
if callers have no need to touch those fields. Since we might well need
to do that again someday in back branches, let's change things so that
only explain.c has to have sizeof(ExplainState) compiled into it. This
costs one extra palloc() per EXPLAIN operation, which is surely pretty
negligible.
As of 9.3, ruleutils.c goes to some lengths to ensure that table and column
aliases used in its output are unique. Of course this takes more time than
was required before, which in itself isn't fatal. However, EXPLAIN was set
up so that recalculation of the unique aliases was repeated for each
subexpression printed in a plan. That results in O(N^2) time and memory
consumption for large plan trees, which did not happen in older branches.
Fortunately, the expensive work is the same across a whole plan tree,
so there is no need to repeat it; we can do most of the initialization
just once per query and re-use it for each subexpression. This buys
back most (not all) of the performance loss since 9.2.
We need an extra ExplainState field to hold the precalculated deparse
context. That's no problem in HEAD, but in the back branches, expanding
sizeof(ExplainState) seems risky because third-party extensions might
have local variables of that struct type. So, in 9.4 and 9.3, introduce
an auxiliary struct to keep sizeof(ExplainState) the same. We should
refactor the APIs to avoid such local variables in future, but that's
material for a separate HEAD-only commit.
Per gripe from Alexey Bashtanov. Back-patch to 9.3 where the issue
was introduced.
Previously, read() might have returned a length equal to the buffer
length, and then the subsequent store to buf[len] would write a
zero-byte one byte past the end. This doesn't seem likely to be
a security issue, but there's some chance it could result in
pg_standby misbehaving.
Spotted by Coverity; patch by Michael Paquier, reviewed by me.
The possibly, depending on compiler settings, generated warning was
"warning: `S_UNLOCK' redefined".
The hppa spinlock implementation doesn't follow the rules of s_lock.h
and provides a gcc specific implementation outside of the the part of
the file that's supposed to do that. It does so to avoid duplication
between the HP compiler and gcc. That unfortunately means that
S_UNLOCK is already defined when the HPPA specific section is reached.
Undefine the generic fallback S_UNLOCK definition inside the HPPA
section. That's far from pretty, but has the big advantage of being
simple. If somebody is interested to fix this in a prettier way...
This presumably got broken in the course of 0709b7ee72.
Discussion: 20150114225919.GY5245@awork2.anarazel.de
Per complaint from Tom Lane.
To do so, move InitializeLatchSupport() into the new common process
initialization functions, and add a new global variable MyLatch.
MyLatch is usable as soon InitPostmasterChild() has been called
(i.e. very early during startup). Initially it points to a process
local latch that exists in all processes. InitProcess/InitAuxiliaryProcess
then replaces that local latch with PGPROC->procLatch. During shutdown
the reverse happens.
This is primarily advantageous for two reasons: For one it simplifies
dealing with the shared process latch, especially in signal handlers,
because instead of having to check for MyProc, MyLatch can be used
unconditionally. For another, a later patch that makes FEs/BE
communication use latches, now can rely on the existence of a latch,
even before having gone through InitProcess.
Discussion: 20140927191243.GD5423@alap3.anarazel.de
Previously, configure would add any switches that it chose of its own
accord to the end of the user-specified CFLAGS string. Since most
compilers process these left-to-right, this meant that configure's choices
would override the user-specified flags in case of conflicts. We'd rather
that worked the other way around, so adjust the logic to put the user's
string at the end not the beginning.
There does not seem to be a need for a similar behavior change for CPPFLAGS
or LDFLAGS: in those, the earlier switches tend to win (think -I or -L
behavior) so putting the user's string at the front is fine.
Backpatch to 9.4 but not earlier. I'm not planning to run buildfarm member
guar on older branches, and it seems a bit risky to change this behavior
in long-stable branches.
Autoconf has known about automatically selecting -Ae when needed for
quite some time now, so remove the redundant addition in template/hpux.
Noted while setting up buildfarm member pademelon.
Since commit 626eb02198 has introduced the auxiliary process
infrastructure, bootstrap_signals() was never used when forked from
postmaster.
Remove the IsUnderPostmaster specific code, and add a appropriate
assertion.
Move common code, that was duplicated in every postmaster child/every
standalone process, into two functions in miscinit.c. Not only does
that already result in a fair amount of net code reduction but it also
makes it much easier to remove more duplication in the future. The
prime motivation wasn't code deduplication though, but easier addition
of new common code.
Commit b94ce6e80 reordered postmaster's startup sequence so that the
tempfile directory is only cleaned up after all the necessary state
for pg_ctl is collected. Unfortunately the chosen location is after
the syslogger has been started; which normally is fine, except for
!WIN32 EXEC_BACKEND builds, which pass information to children via
files in the temp directory.
Move the call to RemovePgTempFiles() to just before the syslogger has
started. That's the first child we fork.
Luckily EXEC_BACKEND is pretty much only used by endusers on windows,
which has a separate method to pass information to children. That
means the real world impact of this bug is very small.
Discussion: 20150113182344.GF12272@alap3.anarazel.de
Backpatch to 9.1, just as the previous commit was.
Since their introduction latches have required barriers in SetLatch
and ResetLatch - but when they were introduced there wasn't any
barrier abstraction. Instead latches were documented to rely on the
callsites to provide barrier semantics.
Now that the barrier support looks halfway complete, add the necessary
barriers to both latch implementations.
Also remove a now superflous lock acquisition from syncrep.c and a
superflous (and insufficient) barrier from freelist.c. There might be
other cases that can now be simplified, but those are the only ones
I've seen on a quick scan.
We might want to backpatch this at some later point, but right now the
barrier infrastructure in the backbranches isn't totally on par with
master.
Discussion: 20150112154026.GB2092@awork2.anarazel.de
So far WaitLatchOrSocket() required to pass in WL_SOCKET_READABLE as
that solely was used to indicate error conditions, like EOF. Waiting
for WL_SOCKET_WRITEABLE would have meant to busy wait upon socket
errors.
Adjust the API to signal errors by returning the socket as readable,
writable or both, depending on WL_SOCKET_READABLE/WL_SOCKET_WRITEABLE
being specified. It would arguably be nicer to return WL_SOCKET_ERROR
but that's not possible on platforms and would probably also result in
more complex callsites.
This previously had explicitly been forbidden in e42a21b9e6, as
there was no strong use case at that point. We now are looking into
making FE/BE communication use latches, so changing this makes sense.
There also are some portability concerns because there cases of older
platforms where select(2) is known to, in violation of POSIX, not
return a socket as writable after the peer has closed it. So far the
platforms where that's the case provide a working poll(2). If we find
one where that's not the case, we'll need to add a workaround for that
platform.
Discussion: 20140927191243.GD5423@alap3.anarazel.de
Reviewed-By: Heikki Linnakangas, Noah Misch
I noticed the "vacuum" regression test taking really significantly longer
than it used to on a slow machine. Investigation pointed the finger at
commit e415b469b3, which added creation of
an index using an extremely expensive index function. That function was
evidently meant to be applied only twice ... but the test re-used an
existing test table, which up till a couple lines before that had had over
two thousand rows. Depending on timing of the concurrent regression tests,
the intervening VACUUMs might have been unable to remove those
recently-dead rows, and then the index build would need to create index
entries for them too, leading to the wrap_do_analyze() function being
executed 2000+ times not twice. Avoid this by using a different table
that is guaranteed to have only the intended two rows in it.
Back-patch to 9.0, like the commit that created the problem.
Some spaces were missing, and putting the affected tuple offset first in
the lock cases instead of the locking data makes more sense.
No backpatch since this is cosmetic and surrounding code has changed.
Commit 3f88672a4 neglected to update the AlterExtensionContentsStmt
production in the grammar to use TypeName to represent types when
passing objects to get_object_address.
Reported as a pg_upgrade failure by Jeff Janes.
The mechanism added in commit dbdf9679d7
for associating the correct translation domain with errcontext strings
potentially fails in cases where errcontext() is used within an ereport()
macro. Such usage was not originally envisioned for errcontext(), but we
do have a few places that do it. In this situation, the intended comma
expression becomes just a couple of arguments to errfinish(), which the
compiler might choose to evaluate right-to-left.
Fortunately, in such cases the textdomain for the errcontext string must
be the same as for the surrounding ereport. So we can fix this by letting
errstart initialize context_domain along with domain; then it will have
the correct value no matter which order the calls occur in. (Note that
error stack callback functions are not invoked until errfinish, so normal
usage of errcontext won't affect what happens for errcontext calls within
the ereport macro.)
In passing, make sure that errcontext calls within the main backend set
context_domain to something non-NULL. This isn't a live bug because
NULL would select the current textdomain() setting which should be the
right thing anyway --- but it seems better to handle this completely
consistently with the regular domain field.
Per report from Dmitry Voronin. Backpatch to 9.3; before that, there
wasn't any attempt to ensure that errcontext strings were translated
in an appropriate domain.
Back in ed0b409, PGPROC was split and moved to static variables in
procarray.c, with procs in ProcArrayStruct replaced by an array of
integers representing process numbers (pgprocnos), with -1 indicating a
dead process which has yet to be removed. Access to procArray is
generally done under ProcArrayLock and therefore most code does not have
to concern itself with -1 entries.
However, MinimumActiveBackends intentionally does not take
ProcArrayLock, which means it has to be extra careful when accessing
procArray. Prior to ed0b409, this was handled by checking for a NULL
in the pointer array, but that check was no longer valid after the
split. Coverity pointed out that the check could never happen and so
it was removed in 5592eba. That didn't make anything worse, but it
didn't fix the issue either.
The correct fix is to check for pgprocno == -1 and skip over that entry
if it is encountered.
Back-patch to 9.2, since there can be attempts to access the arrays
prior to their start otherwise. Note that the changes prior to 9.4 will
look a bit different due to the change in 5592eba.
Note that MinimumActiveBackends only returns a bool for heuristic
purposes and any pre-array accesses are strictly read-only and so there
is no security implication and the lack of fields complaints indicates
it's very unlikely to run into issues due to this.
Pointed out by Noah.
Commit 0eea8047bf introduced some overly
optimistic assumptions about what could be in a local struct variable's
initializer. (This might in fact be valid code according to C99, but I've
got at least one pre-C99 compiler that falls over on those nonconstant
address expressions.) There is no reason whatsoever for main()'s workspace
to not be static, so revert long_options[] to a static and make the
DumpOptions struct static as well.
We had code that supposed that some platforms might offer a nonstandard
version of getpwuid_r() with only four arguments. However, the 5-argument
definition has been standardized at least since the Single Unix Spec v2,
which is our normal reference for what's portable across all Unix-oid
platforms. (What's more, this wasn't the only pre-standardization version
of getpwuid_r(); my old HPUX 10.20 box has still another signature.)
So let's just get rid of the now-useless configure step.
Some users run their applications in chroot environments that lack an
/etc/passwd file. This means that the current UID's user name and home
directory are not obtainable. libpq used to be all right with that,
so long as the database role name to use was specified explicitly.
But commit a4c8f14364 broke such cases by
causing any failure of pg_fe_getauthname() to be treated as a hard error.
In any case it did little to advance its nominal goal of causing errors
in pg_fe_getauthname() to be reported better. So revert that and instead
put some real error-reporting code in place. This requires changes to the
APIs of pg_fe_getauthname() and pqGetpwuid(), since the latter had
departed from the POSIX-specified API of getpwuid_r() in a way that made
it impossible to distinguish actual lookup errors from "no such user".
To allow such failures to be reported, while not failing if the caller
supplies a role name, add a second call of pg_fe_getauthname() in
connectOptions2(). This is a tad ugly, and could perhaps be avoided with
some refactoring of PQsetdbLogin(), but I'll leave that idea for later.
(Note that the complained-of misbehavior only occurs in PQsetdbLogin,
not when using the PQconnect functions, because in the latter we will
never bother to call pg_fe_getauthname() if the user gives a role name.)
In passing also clean up the Windows-side usage of GetUserName(): the
recommended buffer size is 257 bytes, the passed buffer length should
be the buffer size not buffer size less 1, and any error is reported
by GetLastError() not errno.
Per report from Christoph Berg. Back-patch to 9.4 where the chroot
failure case was introduced. The generally poor reporting of errors
here is of very long standing, of course, but given the lack of field
complaints about it we won't risk changing these APIs further back
(even though they're theoretically internal to libpq).
If the compiler/arch combination does not provide compiler barriers,
provide a fallback. That fallback simply consists out of a function
call into a externally defined function. That should guarantee
compiler barrierer semantics except for compilers that do inter
translation unit/global optimization - those better provide an actual
compiler barrier.
Hopefully this fixes Tom's report of linker failures due to
pg_compiler_barrier_impl not being provided.
I'm not backpatching this commit as it builds on the new atomics
infrastructure. If we decide an equivalent fix needs to be
backpatched, I'll do so in a separate commit.
Discussion: 27746.1420930690@sss.pgh.pa.us
Per report from Tom Lane.
I failed to recognize that pg_atomic_uint64 wasn't guaranteed to be 8
byte aligned on some 32bit platforms - which it has to be on some
platforms to guarantee the desired atomicity and which we assert.
As this is all compiler specific code anyway we can just rely on
compiler specific tricks to enforce alignment.
I've been unable to find concrete documentation about the version that
introduce the sunpro alignment support, so that might need additional
guards.
I've verified that this works with gcc x86 32bit, but I don't have
access to any other 32bit environment.
Discussion: op.xpsjdkil0sbe7t@vld-kuci
Per report from Vladimir Koković.
For some reason I overlooked in GETTEXT_TRIGGERS that the right argument
be read by gettext in 7fcbf6a405. This
will drop the translation percentages for the backend all the way back
to 9.3 ...
Problem reported by Heikki.
The event trigger test for rowsecurity can cause problems for other
tests which are run in parallel with it. Instead of running that test
in the rowsecurity set, move it to the event_trigger set, which runs
isolated from other tests.
Also reverts 7161b08, which moved rowsecurity into its own test group.
That's no longer necessary, now that the event trigger test is gone from
the rowsecurity set of tests.
Pointed out by Tom.
The new logging introduced in 35192f06 made the incorrect assumption
that scan_all vacuums would always wait for buffer pins; but they only
do so if the page actually needs to be frozen.
Fix that inaccuracy by removing the difference in log output based on
scan_all and just always remove the same message. I chose to keep the
split log message from the original commit for now, it seems likely
that it'll be of use in the future.
Also merge the line about buffer pins in autovacuum's log output into
the existing "pages: ..." line. It seems odd to have a separate line
about pins, without the "topic: " prefix others have.
Also rename the new 'pinned_pages' variable to 'pinskipped_pages'
because it actually tracks the number of pages that could *not* be
pinned.
Discussion: 20150104005324.GC9626@awork2.anarazel.de
The previous commit introduced its report at LOG level to avoid
surprises at minor release upgrade time. Compel users deploying the
next major release to also deploy the reported workaround.
Darwin --enable-nls builds use a substitute setlocale() that may start a
thread. Buildfarm member orangutan experienced BackendList corruption
on account of different postmaster threads executing signal handlers
simultaneously. Furthermore, a multithreaded postmaster risks undefined
behavior from sigprocmask() and fork(). Emit LOG messages about the
problem and its workaround. Back-patch to 9.0 (all supported versions).
Typical server invocations already achieved that. Invalid locale
settings in the initial postmaster environment interfered, as could
malloc() failure. Setting "LC_MESSAGES=pt_BR.utf8 LC_ALL=invalid" in
the postmaster environment will now choose C-locale messages, not
Brazilian Portuguese messages. Most localized programs, including all
PostgreSQL frontend executables, do likewise. Users are unlikely to
observe changes involving locale categories other than LC_MESSAGES.
CheckMyDatabase() ensures that we successfully set LC_COLLATE and
LC_CTYPE; main() sets the remaining three categories to locale "C",
which almost cannot fail. Back-patch to 9.0 (all supported versions).
vacuum()'s static variable handling makes it non-reentrant; an ensuing
null pointer deference crashed the backend. Back-patch to 9.0 (all
supported versions).
Since commit ba94518a, we used XLogFileOpen to open the next segment for
writing, but if the end-of-recovery happens exactly at a segment boundary,
the new segment might not exist yet. (Before ba94518a, XLogFileOpen was
correct, because we would open the previous segment if the switch happened
at the boundary.)
Instead of trying to create it if necessary, it's simpler to not bother
opening the segment at all. XLogWrite() will open or create it soon anyway,
after writing the checkpoint or end-of-recovery record.
Reported by Andres Freund.
Previously, the xml value resulting from an xpath query would not have
namespace declarations if the namespace declarations were attached to
an ancestor element in the input xml value. That means the output value
was not correct XML. Fix that by running the result value through
xmlCopyNode(), which produces the correct namespace declarations.
Author: Ali Akbar <the.apaan@gmail.com>
When using a historic snapshot for logical decoding it can validly
happen that a relation that's in the relcache isn't visible to that
historic snapshot. E.g. if a newly created relation is referenced in
the query that uses the SQL interface for logical decoding and a
sinval reset occurs.
The earlier commit that fixed the error handling for that corner case
already improves the situation as a ERROR is better than hitting an
assertion... But it's obviously not good enough. So additionally
allow that case without an error if a historic snapshot is set up -
that won't allow an invalid entry to stay in the cache because it's a)
already marked invalid and will thus be rebuilt during the next access
b) the syscaches will be reset at the end of decoding.
There might be prettier solutions to handle this case, but all that we
could think of so far end up being much more complex than this quite
simple fix.
This fixes the assertion failures reported by the buildfarm (markhor,
tick, leech) after the introduction of new regression tests in
89fd41b390. The failure there weren't actually directly caused by
CLOBBER_CACHE_ALWAYS but the extraordinary long runtimes due to it
lead to sinval resets triggering the behaviour.
Discussion: 22459.1418656530@sss.pgh.pa.us
Backpatch to 9.4 where logical decoding was introduced.
The corner case where a relcache invalidation tried to rebuild the
entry for a referenced relation but couldn't find it in the catalog
wasn't correct.
The code tried to RelationCacheDelete/RelationDestroyRelation the
entry. That didn't work when assertions are enabled because the latter
contains an assertion ensuring the refcount is zero. It's also more
generally a bad idea, because by virtue of being referenced somebody
might actually look at the entry, which is possible if the error is
trapped and handled via a subtransaction abort.
Instead just error out, without deleting the entry. As the entry is
marked invalid, the worst that can happen is that the invalid (and at
some point unused) entry lingers in the relcache.
Discussion: 22459.1418656530@sss.pgh.pa.us
There should be no way to hit this case < 9.4 where logical decoding
introduced a bug that can hit this. But since the code for handling
the corner case is there it should do something halfway sane, so
backpatch all the the way back. The logical decoding bug will be
handled in a separate commit.