Tom Lane [Tue, 7 Mar 2017 15:42:11 +0000 (10:42 -0500)]
Remove vestigial grammar support for CHARACTER ... CHARACTER SET option.
The SQL standard says that you should be able to write "CHARACTER SET foo"
as part of the declaration of a char-type column. We don't implement that,
but a rough form of support has existed in gram.y since commit f10b63923.
That's now sat there for nigh 20 years without anyone fleshing it out ---
and even if someone did, the contemplated approach of having separate data
type name(s) for every character set certainly isn't what we'd do today.
Let's just remove the grammar production; if anyone is ever motivated to
work on this, reinventing the grammar support is a trivial fraction of
what they'd have to do. And we've never documented anything about
supporting such a clause.
Robert Haas [Tue, 7 Mar 2017 15:33:29 +0000 (10:33 -0500)]
Preparatory refactoring for parallel merge join support.
Extract the logic used by hash_inner_and_outer into a separate
function, get_cheapest_parallel_safe_total_inner, so that it can
also be used to plan parallel merge joins.
Also, add a require_parallel_safe argument to the existing function
get_cheapest_path_for_pathkeys, because parallel merge join needs
to find the cheapest path for a given set of pathkeys that is
parallel-safe, not just the cheapest one overall.
Robert Haas [Tue, 7 Mar 2017 15:22:07 +0000 (10:22 -0500)]
Fix parallel hash join path search.
When the very cheapest path is not parallel-safe, we want to instead use
the cheapest unparameterized path that is. The old code searched
innerrel->cheapest_parameterized_paths, but that isn't right, because
the path we want may not be in that list. Search innerrel->pathlist
instead.
Stephen Frost [Tue, 7 Mar 2017 14:31:52 +0000 (09:31 -0500)]
psql: Add \gx command
It can often be useful to use expanded mode output (\x) for just a
single query. Introduce a \gx which acts exactly like \g except that it
will force expanded output mode for that one \gx call. This is simpler
than having to use \x as a toggle and also means that the user doesn't
have to worry about the current state of the expanded variable, or
resetting it later, to ensure a given query is always returned in
expanded mode.
Primairly Christoph's patch, though I did tweak the documentation and help
text a bit, and re-indented the tab completion section.
Author: Christoph Berg
Reviewed By: Daniel Verite
Discussion: https://postgr.es/m/20170127132737.6skslelaf4txs6iw%40msg.credativ.de
Simon Riggs [Tue, 7 Mar 2017 14:00:54 +0000 (22:00 +0800)]
Allow pg_dumpall to dump roles w/o user passwords
Add new option --no-role-passwords which dumps roles without passwords.
Since we don’t need passwords, we choose to use pg_roles in preference
to pg_authid since access may be restricted for security reasons in
some configrations.
Support SCRAM-SHA-256 authentication (RFC 5802 and 7677).
This introduces a new generic SASL authentication method, similar to the
GSS and SSPI methods. The server first tells the client which SASL
authentication mechanism to use, and then the mechanism-specific SASL
messages are exchanged in AuthenticationSASLcontinue and PasswordMessage
messages. Only SCRAM-SHA-256 is supported at the moment, but this allows
adding more SASL mechanisms in the future, without changing the overall
protocol.
Support for channel binding, aka SCRAM-SHA-256-PLUS is left for later.
The SASLPrep algorithm, for pre-processing the password, is not yet
implemented. That could cause trouble, if you use a password with
non-ASCII characters, and a client library that does implement SASLprep.
That will hopefully be added later.
Authorization identities, as specified in the SCRAM-SHA-256 specification,
are ignored. SET SESSION AUTHORIZATION provides more or less the same
functionality, anyway.
If a user doesn't exist, perform a "mock" authentication, by constructing
an authentic-looking challenge on the fly. The challenge is derived from
a new system-wide random value, "mock authentication nonce", which is
created at initdb, and stored in the control file. We go through these
motions, in order to not give away the information on whether the user
exists, to unauthenticated users.
Bumps PG_CONTROL_VERSION, because of the new field in control file.
Patch by Michael Paquier and Heikki Linnakangas, reviewed at different
stages by Robert Haas, Stephen Frost, David Steele, Aleksander Alekseev,
and many others.
Stephen Frost [Tue, 7 Mar 2017 04:29:02 +0000 (23:29 -0500)]
pg_dump: Properly handle public schema ACLs with --clean
pg_dump has always handled the public schema in a special way when it
comes to the "--clean" option. To wit, we do not drop or recreate the
public schema in "normal" mode, but when we are run in "--clean" mode
then we do drop and recreate the public schema.
When running in "--clean" mode, the public schema is dropped and then
recreated and it is recreated with the normal schema-default privileges
of "nothing". This is unlike how the public schema starts life, which
is to have CREATE and USAGE GRANT'd to the PUBLIC role, and that is what
is recorded in pg_init_privs.
Due to this, in "--clean" mode, pg_dump would mistakenly only dump out
the set of privileges required to go from the initdb-time privileges on
the public schema to whatever the current-state privileges are. If the
privileges were not changed from initdb time, then no privileges would
be dumped out for the public schema, but with the schema being dropped
and recreated, the result was that the public schema would have no ACLs
on it instead of what it should have, which is the initdb-time
privileges.
Practically speaking, this meant that pg_dump with --clean mode dumping
a database where the ACLs on the public schema were not changed from the
default would, upon restore, result in a public schema with *no*
privileges GRANT'd, not matching the state of the existing database
(where the initdb-time privileges would have been CREATE and USAGE to
the PUBLIC role for the public schema).
To fix, adjust the query in getNamespaces() to ignore the pg_init_privs
entry for the public schema when running in "--clean" mode, meaning that
the privileges for the public schema would be dumped, correctly, as if
it was going from a newly-created schema to the current state (which is,
indeed, what will happen during the restore thanks to the DROP/CREATE).
Only the public schema is handled in this special way by pg_dump, no
other initdb-time objects are dropped/recreated in --clean mode.
Tom Lane [Tue, 7 Mar 2017 00:33:59 +0000 (19:33 -0500)]
Repair incorrect pg_dump labeling for some comments and security labels.
We attached no schema label to comments for procedural languages, casts,
transforms, operator classes, operator families, or text search objects.
The first three categories of objects don't really have schemas, but
pg_dump treats them as if they do, and it seems like the TocEntry fields
for their comments had better match the TocEntry fields for the parent
objects. (As an example of a possible hazard, the type names in a CAST
will be formatted with the assumption of a particular search_path, so
failing to ensure that this same path is active for the COMMENT ON command
could lead to an error or to attaching the comment to the wrong cast.)
In the last six cases, this was a flat-out error --- possibly mine to
begin with, but it was a long time ago.
The security label for a procedural language was likewise not correctly
labeled as to schema, and both the comment and security label for a
procedural language were not correctly labeled as to owner.
In simple cases the restore would accidentally work correctly anyway, since
these comments and security labels would normally get emitted right after
the owning object, and so the search path and active user would be correct
anyhow. But it could fail in corner cases; for example a schema-selective
restore would omit comments it should include.
Giuseppe Broccolo noted the oversight, and proposed the correct fix, for
text search dictionary objects; I found the rest by cross-checking other
dumpComment() calls. These oversights are ancient, so back-patch all
the way.
Andres Freund [Wed, 23 Nov 2016 08:23:42 +0000 (00:23 -0800)]
Make simplehash.h grow hashtable in additional cases.
Increase the size when either the distance between actual and optimal
slot grows too large, or when too many subsequent entries would have
to be moved.
This addresses reports that the simplehash performed, sometimes
considerably, worse than dynahash.
The reason turned out to be that insertions into the hashtable where,
due to the use of parallel query, in effect done from another
hashtable, in hash-value order. If the target hashtable, due to
mis-estimation, was sized a lot smaller than the source table(s) that
lead to very imbalanced tables; a lot of entries in many close-by
buckets from the source tables were inserted into a single, wider,
bucket on the target table. As the growth factor was solely computed
based on the fillfactor, the performance of the table decreased
further and further.
b81b5a96f424531b was an attempt to address this problem for hash
aggregates (but not for bitmap scans), but it turns out that the
current method of mixing hash values often actually leaves neighboring
hash-values close to each other, just in different value range. It
might be worth revisiting that independently of the performance issues
addressed in this patch..
To address that problem resize tables in two additional cases: Firstly
when the optimal position for an entry would be far from the actual
position, secondly when many entries would have to be moved to make
space for the new entry (while satisfying the robin hood property).
Due to the additional resizing threshold it seems possible, and
testing confirms that so far, that a higher fillfactor doesn't hurt
performance and saves a bit of memory. It seems better to increase it
now, before a release containing any of this code, rather than wonder
in some later release.
The various boundaries aren't determined in a particularly scientific
manner, they might need some fine-tuning.
In all my tests the new code now, even with parallelism, performs at
least as good as the old code, in several scenarios significantly
better.
Reported-By: Dilip Kumar, Robert Haas, Kuntal Ghosh
Discussion:
https://postgr.es/m/CAFiTN-vagvuAydKG9VnWcoK=ADAhxmOa4ZTrmNsViBBooTnriQ@mail.gmail.com
https://postgr.es/m/CAGz5QC+=fNTYgzMLTBUNeKt6uaWZFXJbkB5+7oWm-n9DwVxcLA@mail.gmail.com
Stephen Frost [Mon, 6 Mar 2017 22:03:57 +0000 (17:03 -0500)]
pg_upgrade: Fix large object COMMENTS, SECURITY LABELS
When performing a pg_upgrade, we copy the files behind pg_largeobject
and pg_largeobject_metadata, allowing us to avoid having to dump out and
reload the actual data for large objects and their ACLs.
Unfortunately, that isn't all of the information which can be associated
with large objects. Currently, we also support COMMENTs and SECURITY
LABELs with large objects and these were being silently dropped during a
pg_upgrade as pg_dump would skip everything having to do with a large
object and pg_upgrade only copied the tables mentioned to the new
cluster.
As the file copies happen after the catalog dump and reload, we can't
simply include the COMMENTs and SECURITY LABELs in pg_dump's binary-mode
output but we also have to include the actual large object definition as
well. With the definition, comments, and security labels in the pg_dump
output and the file copies performed by pg_upgrade, all of the data and
metadata associated with large objects is able to be successfully pulled
forward across a pg_upgrade.
In 9.6 and master, we can simply adjust the dump bitmask to indicate
which components we don't want. In 9.5 and earlier, we have to put
explciit checks in in dumpBlob() and dumpBlobs() to not include the ACL
or the data when in binary-upgrade mode.
Adjustments made to the privileges regression test to allow another test
(large_object.sql) to be added which explicitly leaves a large object
with a comment in place to provide coverage of that case with
pg_upgrade.
Tom Lane [Mon, 6 Mar 2017 21:50:47 +0000 (16:50 -0500)]
Avoid dangling pointer to relation name in RLS code path in DoCopy().
With RLS active, "COPY tab TO ..." failed under -DRELCACHE_FORCE_RELEASE,
and would sometimes fail without that, because it used the relation name
directly from the relcache as part of the parsetree it's building. That
becomes a potentially-dangling pointer as soon as the relcache entry is
closed, a bit further down. Typical symptom if the relcache entry chanced
to get cleared would be "relation does not exist" error with a garbage
relation name, or possibly a core dump; but if you were really truly
unlucky, the COPY might copy from the wrong table.
Per report from Andrew Dunstan that regression tests fail with
-DRELCACHE_FORCE_RELEASE. The core tests now pass for me (but have
not tried "make check-world" yet).
Peter Eisentraut [Wed, 28 Dec 2016 17:00:00 +0000 (12:00 -0500)]
Allow dropping multiple functions at once
The generic drop support already supported dropping multiple objects of
the same kind at once. But the previous representation
of function signatures across two grammar symbols and structure members
made this cumbersome to do for functions, so it was not supported. Now
that function signatures are represented by a single structure, it's
trivial to add this support. Same for aggregates and operators.
Reviewed-by: Jim Nasby <Jim.Nasby@BlueTreble.com> Reviewed-by: Michael Paquier <michael.paquier@gmail.com>
Peter Eisentraut [Wed, 28 Dec 2016 17:00:00 +0000 (12:00 -0500)]
Replace LookupFuncNameTypeNames() with LookupFuncWithArgs()
The old function took function name and function argument list as
separate arguments. Now that all function signatures are passed around
as ObjectWithArgs structs, this is no longer necessary and can be
replaced by a function that takes ObjectWithArgs directly. Similarly
for aggregates and operators.
Reviewed-by: Jim Nasby <Jim.Nasby@BlueTreble.com> Reviewed-by: Michael Paquier <michael.paquier@gmail.com>
Peter Eisentraut [Sat, 12 Nov 2016 17:00:00 +0000 (12:00 -0500)]
Remove objname/objargs split for referring to objects
In simpler times, it might have worked to refer to all kinds of objects
by a list of name components and an optional argument list. But this
doesn't work for all objects, which has resulted in a collection of
hacks to place various other nodes types into these fields, which have
to be unpacked at the other end. This makes it also weird to represent
lists of such things in the grammar, because they would have to be lists
of singleton lists, to make the unpacking work consistently. The other
problem is that keeping separate name and args fields makes it awkward
to deal with lists of functions.
Change that by dropping the objargs field and have objname, renamed to
object, be a generic Node, which can then be flexibly assigned and
managed using the normal Node mechanisms. In many cases it will still
be a List of names, in some cases it will be a string Value, for types
it will be the existing Typename, for functions it will now use the
existing ObjectWithArgs node type. Some of the more obscure object
types still use somewhat arbitrary nested lists.
Reviewed-by: Jim Nasby <Jim.Nasby@BlueTreble.com> Reviewed-by: Michael Paquier <michael.paquier@gmail.com>
Robert Haas [Mon, 6 Mar 2017 18:10:55 +0000 (13:10 -0500)]
Fix incorrect comments.
Commit 19dc233c32f2900e57b8da4f41c0f662ab42e080 introduced these
comments. Michael Paquier noticed that one of them had a typo, but
a bigger problem is that they were not an accurate description of
what the code was doing.
Robert Haas [Mon, 6 Mar 2017 17:41:55 +0000 (12:41 -0500)]
Mark pg_start_backup and pg_stop_backup as parallel-restricted.
They depend on backend-private state that will not be synchronized by
the parallel machinery, so they should not be marked parallel-safe.
This issue also exists in 9.6, but we obviously can't do anything
about 9.6 clusters that already exist. Possibly this could be
back-patched so that future 9.6 clusters would come out OK, or
possibly we should back-patch some other fix, but that would need more
discussion.
Simon Riggs [Mon, 6 Mar 2017 10:34:31 +0000 (16:04 +0530)]
Reduce lock levels for table storage params related to planning
The following parameters are now updateable with ShareUpdateExclusiveLock
effective_io_concurrency
parallel_workers
seq_page_cost
random_page_cost
n_distinct
n_distinct_inherited
Tom Lane [Sat, 4 Mar 2017 21:09:33 +0000 (16:09 -0500)]
In rebuild_relation(), don't access an already-closed relcache entry.
This reliably fails with -DRELCACHE_FORCE_RELEASE, as reported by
Andrew Dunstan, and could sometimes fail in normal operation, resulting
in a wrong persistence value being used for the transient table.
It's not immediately clear to me what effects that might have beyond
the risk of a crash while accessing OldHeap->rd_rel->relpersistence,
but it's probably not good.
Bug introduced by commit f41872d0c, and made substantially worse by
commit 85b506bbf, which added a second such access significantly
later than the heap_close. I doubt the first reference could fail
in a production scenario, but the second one definitely could.
Disallow CREATE/DROP SUBSCRIPTION in transaction block
Disallow CREATE SUBSCRIPTION and DROP SUBSCRIPTION in a transaction
block when the replication slot is to be created or dropped, since that
cannot be rolled back.
based on patch by Masahiko Sawada <sawada.mshk@gmail.com>
Add tab completion for publications and subscriptions. Also, to be able
to get a list of subscriptions, make pg_subscription world-readable but
revoke access to subconninfo using column privileges.
This makes the connection attempt from CREATE SUBSCRIPTION and from
WalReceiver interruptable by the user in case the libpq connection is
hanging. The previous coding required immediate shutdown (SIGQUIT) of
PostgreSQL in that situation.
From: Petr Jelinek <petr.jelinek@2ndquadrant.com> Tested-by: Thom Brown <thom@linux.com>
Simon Riggs [Fri, 3 Mar 2017 13:48:25 +0000 (19:18 +0530)]
Allow vacuums to report oldestxmin
Allow VACUUM and Autovacuum to report the oldestxmin value they
used while cleaning tables, helping to make better sense out of
the other statistics we report in various cases.
Robert Haas [Fri, 3 Mar 2017 06:02:45 +0000 (11:32 +0530)]
Add pg_current_logfile() function.
The syslogger will write out the current stderr and csvlog names, if
it's running and there are any, to a new file in the data directory
called "current_logfiles". We take care to remove this file when it
might no longer be valid (but not at shutdown). The function
pg_current_logfile() can be used to read the entries in the file.
Gilles Darold, reviewed and modified by Karl O. Pinc, Michael
Paquier, and me. Further review by Álvaro Herrera and Christoph Berg.
Robert Haas [Fri, 3 Mar 2017 03:44:49 +0000 (09:14 +0530)]
Notify bgworker registrant after freeing worker slot.
Tom Lane observed buildfarm failures caused by the select_parallel
regression test trying to launch new parallel queries before the
worker slots used by the previous ones were freed. Try to fix this by
having the postmaster free the worker slots before it sends the
SIGUSR1 notifications to the registering process. This doesn't
completely eliminate the possibility that the user backend might
(correctly) observe the worker as dead before the slot is free, but I
believe it should make the window significantly narrower.
Patch by me, per complaint from Tom Lane. Reviewed by Amit Kapila.
Robert Haas [Fri, 3 Mar 2017 03:37:41 +0000 (09:07 +0530)]
Improve error reporting for tuple-routing failures.
Currently, the whole row is shown without column names. Instead,
adopt a style similar to _bt_check_unique() in ExecFindPartition()
and show the failing key: (key1, ...) = (val1, ...).
Amit Langote, per a complaint from Simon Riggs. Reviewed by me;
I also adjusted the grammar in one of the comments.
Robert Haas [Thu, 2 Mar 2017 13:17:40 +0000 (18:47 +0530)]
Refactor bitmap heap scan in preparation for parallel support.
The final patch will be less messy if the prefetching support is
a bit better isolated, so do that.
Dilip Kumar, with some changes by me. The larger patch set of which
this is a part has been reviewed and tested by (at least) Andres
Freund, Amit Khandekar, Tushar Ahuja, Rafia Sabih, Haribabu Kommi, and
Thomas Munro.
Robert Haas [Thu, 2 Mar 2017 11:48:19 +0000 (17:18 +0530)]
Don't uselessly rewrite, truncate, VACUUM, or ANALYZE partitioned tables.
Also, recursively perform VACUUM and ANALYZE on partitions when the
command is applied to a partitioned table. In passing, some related
documentation updates.
Amit Langote, reviewed by Michael Paquier, Ashutosh Bapat, and by me.
Noah Misch [Thu, 2 Mar 2017 05:03:27 +0000 (00:03 -0500)]
Handle unaligned SerializeSnapshot() buffer.
Likewise in RestoreSnapshot(). Do so by copying between the user buffer
and a stack buffer of known alignment. Back-patch to 9.6, where this
last applies cleanly. In master, the select_parallel test dies with
SIGBUS on "Oracle Solaris 10 1/13 s10s_u11wos_24a SPARC", building
32-bit with gcc 4.9.2. In 9.6 and 9.5, the buffers in question happen
to be sufficiently-aligned, and this change is mere insurance against
future 9.6 changes or extension code compromising that.
Amit Kapila, with slight changes by me. The larger patch series of
which this is a part has been reviewed and tested by Álvaro Herrera,
Ashutosh Sharma, Mark Kirkwood, Jeff Janes, and Jesper Pedersen.
Andres Freund [Wed, 1 Mar 2017 04:41:44 +0000 (20:41 -0800)]
Fix assertion failure due to over-eager code deduplication.
In the previous commit I'd made MemoryContextContains() use
GetMemoryChunkContext(), but that causes trouble when the passed
pointer isn't allocated in any memory context - that's probably
something we shouldn't do, but the previous commit isn't a place for a
"policy" change.
Andres Freund [Tue, 28 Feb 2017 18:36:29 +0000 (10:36 -0800)]
Overhaul memory management README.
The README was written as a "historical account", and that style
hasn't aged particularly well. Rephrase it to describe the current
situation, instead of having various version specific comments.
This also updates the description of how allocated chunks are
associated with their corresponding context, the method of which has
changed in the preceding commit.
Author: Andres Freund
Discussion: https://postgr.es/m/20170228074420.aazv4iw6k562mnxg@alap3.anarazel.de
Andres Freund [Tue, 28 Feb 2017 07:32:22 +0000 (23:32 -0800)]
Reduce size of common allocation header.
The new slab allocator needs different per-allocation information than
the classical aset.c. The definition in 58b25e981 wasn't sufficiently
careful on 32 platforms with 8 byte alignment, leading to buildfarm
failures. That's not entirely easy to fix by just adjusting the
definition.
As slab.c doesn't actually need the size part(s) of the common header,
all chunks are equally sized after all, it seems better to instead
reduce the header to the part needed by all allocators, namely which
context an allocation belongs to. That has the advantage of reducing
the overhead of slab allocations, and also allows for more flexibility
in future allocators.
To avoid spreading the logic about accessing a chunk's context around,
centralize it in GetMemoryChunkContext(), which allows to delete a
good number of lines.
A followup commit will revise the mmgr/README portion about
StandardChunkHeader, and more.
Author: Andres Freund
Discussion: https://postgr.es/m/20170228074420.aazv4iw6k562mnxg@alap3.anarazel.de
Tom Lane [Mon, 27 Feb 2017 22:20:34 +0000 (17:20 -0500)]
Allow index AMs to return either HeapTuple or IndexTuple format during IOS.
Previously, only IndexTuple format was supported for the output data of
an index-only scan. This is fine for btree, which is just returning a
verbatim index tuple anyway. It's not so fine for SP-GiST, which can
return reconstructed data that's much larger than a page.
To fix, extend the index AM API so that index-only scan data can be
returned in either HeapTuple or IndexTuple format. There's other ways
we could have done it, but this way avoids an API break for index AMs
that aren't concerned with the issue, and it costs little except a couple
more fields in IndexScanDescs.
I changed both GiST and SP-GiST to use the HeapTuple method. I'm not
very clear on whether GiST can reconstruct data that's too large for an
IndexTuple, but that seems possible, and it's not much of a code change to
fix.
Per a complaint from Vik Fearing. Reviewed by Jason Li.
Robert Haas [Mon, 27 Feb 2017 17:26:34 +0000 (22:56 +0530)]
hash: Refactor overflow page allocation.
As with commit b0f18cb77f50a54e997d857d592f6a511617f52c, the goal
here is to move all of the related page modifications to a single
section of code, in preparation for adding write-ahead logging.
Amit Kapila, with slight changes by me. The larger patch series
of which this is a part has been reviewed and tested by Álvaro
Herrera, Ashutosh Sharma, Mark Kirkwood, Jeff Janes, and Jesper
Pedersen, all of whom should also have been credited in the
previous commit message.
Robert Haas [Mon, 27 Feb 2017 17:04:21 +0000 (22:34 +0530)]
hash: Refactor bucket squeeze code.
In preparation for adding write-ahead logging to hash indexes,
refactor _hash_freeovflpage and _hash_squeezebucket so that all
related page modifications happen in a single section of code. The
previous coding assumed that it would be fine to move tuples one at a
time, and also that the various operations involved in freeing an
overflow page didn't necessarily all need to be done together, all
of which is true if you don't care about write-ahead logging.
Tom Lane [Mon, 27 Feb 2017 16:20:22 +0000 (11:20 -0500)]
Remove PL/Tcl's "module" facility.
PL/Tcl has long had a facility whereby Tcl code could be autoloaded from
a database table named "pltcl_modules". However, nobody is using it, as
evidenced by the recent discovery that it's never been fixed to work with
standard_conforming_strings turned on. Moreover, it's rather shaky from
a security standpoint, and the table design is very old and crufty (partly
because it dates from before we had TOAST). A final problem is that
because the table-population scripts depend on the Tcl client library
Pgtcl, which we removed from the core distribution in 2004, it's
impossible to create a self-contained regression test for the feature.
Rather than try to surmount these problems, let's just remove it.
A follow-on patch will provide a way to execute user-defined
initialization code, similar to features that exist in plperl and plv8.
With that, it will be possible to implement this feature or similar ones
entirely in userspace, which is where it belongs.
Peter Eisentraut [Mon, 27 Feb 2017 13:30:06 +0000 (08:30 -0500)]
chomp PQerrorMessage() in backend uses
PQerrorMessage() returns an error message with a trailing newline, but
in backend use (dblink, postgres_fdw, libpqwalreceiver), we want to have
the error message without that for emitting via ereport(). To simplify
that, add a function pchomp() that returns a pstrdup'ed string with the
trailing newline characters removed.
Andres Freund [Mon, 27 Feb 2017 11:41:44 +0000 (03:41 -0800)]
Use the new "Slab" context for some allocations in reorderbuffer.h.
Note that this change alone does not yet fully address the performance
problems triggering this work, a large portion of the slowdown is
triggered by the tuple allocator, which isn't converted to the new
allocator. It would be possible to do so, but using evenly sized
objects, like both the current implementation in reorderbuffer.c and
slab.c, wastes a fair amount of memory. A later patch by Tomas will
introduce a better approach.
Author: Tomas Vondra Reviewed-By: Andres Freund
Discussion: https://postgr.es/m/d15dff83-0b37-28ed-0809-95a5cc7292ad@2ndquadrant.com
Andres Freund [Mon, 27 Feb 2017 11:41:44 +0000 (03:41 -0800)]
Add "Slab" MemoryContext implementation for efficient equal-sized allocations.
The default general purpose aset.c style memory context is not a great
choice for allocations that are all going to be evenly sized,
especially when those objects aren't small, and have varying
lifetimes. There tends to be a lot of fragmentation, larger
allocations always directly go to libc rather than have their cost
amortized over several pallocs.
These problems lead to the introduction of ad-hoc slab allocators in
reorderbuffer.c. But it turns out that the simplistic implementation
leads to problems when a lot of objects are allocated and freed, as
aset.c is still the underlying implementation. Especially freeing can
easily run into O(n^2) behavior in aset.c.
While the O(n^2) behavior in aset.c can, and probably will, be
addressed, custom allocators for this behavior are more efficient
both in space and time.
This allocator is for evenly sized allocations, and supports both
cheap allocations and freeing, without fragmenting significantly. It
does so by allocating evenly sized blocks via malloc(), and carves
them into chunks that can be used for allocations. In order to
release blocks to the OS as early as possible, chunks are allocated
from the fullest block that still has free objects, increasing the
likelihood of a block being entirely unused.
A subsequent commit uses this in reorderbuffer.c, but a further
allocator is needed to resolve the performance problems triggering
this work.
There likely are further potentialy uses of this allocator besides
reorderbuffer.c.
There's potential further optimizations of the new slab.c, in
particular the array of freelists could be replaced by a more
intelligent structure - but for now this looks more than good enough.
Author: Tomas Vondra, editorialized by Andres Freund Reviewed-By: Andres Freund, Petr Jelinek, Robert Haas, Jim Nasby
Discussion: https://postgr.es/m/d15dff83-0b37-28ed-0809-95a5cc7292ad@2ndquadrant.com
Andres Freund [Mon, 27 Feb 2017 11:41:44 +0000 (03:41 -0800)]
Make useful infrastructure from aset.c generally available.
An upcoming patch introduces a new type of memory context. To avoid
duplicating debugging infrastructure within aset.c, move useful pieces
to memdebug.[ch].
While touching aset.c, fix printf format code in AllocFree* debug
macros.
Author: Tomas Vondra Reviewed-By: Andres Freund
Discussion: https://postgr.es/m/b3b2245c-b37a-e1e5-ebc4-857c914bc747@2ndquadrant.com
Robert Haas [Sun, 26 Feb 2017 08:06:49 +0000 (13:36 +0530)]
Allow custom and foreign scans to have shutdown callbacks.
This is expected to be useful mostly when performing such scans in
parallel, because in that case it allows (in combination with commit acf555bc53acb589b5a2827e65d655fa8c9adee0) nodes below a Gather to get
control just before the DSM segment goes away.
KaiGai Kohei, except that I rewrote the documentation. Reviewed by
Claudio Freire.
Tom Lane [Sat, 25 Feb 2017 23:34:09 +0000 (18:34 -0500)]
Put back #include <windows.h> in dirmod.c.
I removed this in commit 9e3755ecb, reasoning that the win32.h
port-specific header file included by c.h would have provided it.
However, that's only true on native win32 builds, not Cygwin builds.
It may be that some of the other <windows.h> inclusions also need
to be put back on the same grounds; but this is the only one that
is clearly meant to be included #ifdef __CYGWIN__, so maybe this is
the extent of the problem. Awaiting further buildfarm results.
Tom Lane [Sat, 25 Feb 2017 23:10:09 +0000 (18:10 -0500)]
Remove some configure header-file checks that we weren't really using.
We had some AC_CHECK_HEADER tests that were really wastes of cycles,
because the code proceeded to #include those headers unconditionally
anyway, in all or a large majority of cases. The lack of complaints
shows that those headers are available on every platform of interest,
so we might as well let configure run a bit faster by not probing
those headers at all.
I suspect that some of the tests I left alone are equally useless, but
since all the existing #includes of the remaining headers are properly
guarded, I didn't touch them.
Tom Lane [Sat, 25 Feb 2017 21:12:24 +0000 (16:12 -0500)]
Remove useless duplicate inclusions of system header files.
c.h #includes a number of core libc header files, such as <stdio.h>.
There's no point in re-including these after having read postgres.h,
postgres_fe.h, or c.h; so remove code that did so.
While at it, also fix some places that were ignoring our standard pattern
of "include postgres[_fe].h, then system header files, then other Postgres
header files". While there's not any great magic in doing it that way
rather than system headers last, it's silly to have just a few files
deviating from the general pattern. (But I didn't attempt to enforce this
globally, only in files I was touching anyway.)
I'd be the first to say that this is mostly compulsive neatnik-ism,
but over time it might save enough compile cycles to be useful.
Bruce Momjian [Sat, 25 Feb 2017 17:59:23 +0000 (12:59 -0500)]
pg_upgrade docs: clarify instructions on standby extensions
Previously the pg_upgrade standby upgrade instructions said not to
execute pgcrypto.sql, but it should have referenced the extension
command "CREATE EXTENSION pgcrypto". This patch makes that doc change.
Reported-by: a private bug report
Backpatch-through: 9.4, where standby instructions were added
Tom Lane [Fri, 24 Feb 2017 21:45:32 +0000 (16:45 -0500)]
Suppress compiler warnings in ecpg test on newer Windows toolchains.
nan_test.pgc supposed that it could unconditionally #define isnan()
and isinf() on WIN32. This was evidently copied at some point from
src/include/port/win32.h, but nowadays there's a test on _MSC_VER
there. Make nan_test.pgc look the same.
Per buildfarm warnings. There's no evidence this produces anything
worse than a warning, and besides it's only a test case, so I don't
feel a need to back-patch.
Tom Lane [Fri, 24 Feb 2017 20:21:39 +0000 (15:21 -0500)]
Fix unportable definition of BSWAP64() macro.
We have a portable way of writing uint64 constants, but whoever wrote
this macro didn't know about it.
While at it, fix unsafe under-parenthesization of arguments. That might
be moot, because there are already good reasons not to use the macro on
anything more complicated than a simple variable, but it's still poor
practice.
Tom Lane [Fri, 24 Feb 2017 03:08:10 +0000 (22:08 -0500)]
Add an Assert that enum_cmp_internal() gets passed an FmgrInfo pointer.
If someone were to try to call one of the enum comparison functions
using DirectFunctionCallN, it would very likely seem to work, because
only in unusual cases does enum_cmp_internal() need to access the
typcache. But once such a case occurred, code like that would crash
with a null pointer dereference. To make an oversight of that sort
less likely to escape detection, add a non-bypassable Assert that
fcinfo->flinfo isn't NULL.
Tom Lane [Thu, 23 Feb 2017 20:57:08 +0000 (15:57 -0500)]
Consistently declare timestamp variables as TimestampTz.
Twiddle the replication-related code so that its timestamp variables
are declared TimestampTz, rather than the uninformative "int64" that
was previously used for meant-to-be-always-integer timestamps.
This resolves the int64-vs-TimestampTz declaration inconsistencies
introduced by commit 7c030783a, though in the opposite direction to
what was originally suggested.
This required including datatype/timestamp.h in a couple more places
than before. I decided it would be a good idea to slim down that
header by not having it pull in <float.h> etc, as those headers are
no longer at all relevant to its purpose. Unsurprisingly, a small number
of .c files turn out to have been depending on those inclusions, so add
them back in the .c files as needed.
Tom Lane [Thu, 23 Feb 2017 17:23:12 +0000 (12:23 -0500)]
Remove pg_control's enableIntTimes field.
We don't need it any more.
pg_controldata continues to report that date/time type storage is
"64-bit integers", but that's now a hard-wired behavior not something
it sees in the data. This avoids breaking pg_upgrade, and perhaps other
utilities that inspect pg_control this way. Ditto for pg_resetwal.
I chose to remove the "bigint_timestamps" output column of
pg_control_init(), though, as that function hasn't been around long
and probably doesn't have ossified users.
Tom Lane [Thu, 23 Feb 2017 16:40:12 +0000 (11:40 -0500)]
De-support floating-point timestamps.
Per discussion, the time has come to do this. The handwriting has been
on the wall at least since 9.0 that this would happen someday, whenever
it got to be too much of a burden to support the float-timestamp option.
The triggering factor now is the discovery that there are multiple bugs
in the code that attempts to implement use of integer timestamps in the
replication protocol even when the server is built for float timestamps.
The internal float timestamps leak into the protocol fields in places.
While we could fix the identified bugs, there's a very high risk of
introducing more. Trying to build a wall that would positively prevent
mixing integer and float timestamps is more complexity than we want to
undertake to maintain a long-deprecated option. The fact that these
bugs weren't found through testing also indicates a lack of interest
in float timestamps.
This commit disables configure's --disable-integer-datetimes switch
(it'll still accept --enable-integer-datetimes, though), removes direct
references to USE_INTEGER_DATETIMES, and removes discussion of float
timestamps from the user documentation. A considerable amount of code is
rendered dead by this, but removing that will occur as separate mop-up.
Tom Lane [Wed, 22 Feb 2017 20:04:07 +0000 (15:04 -0500)]
Fix contrib/pg_trgm's extraction of trigrams from regular expressions.
The logic for removing excess trigrams from the result was faulty.
It intends to avoid merging the initial and final states of the NFA,
which is necessary, but in testing whether removal of a specific trigram
would cause that, it failed to consider the combined effects of all the
state merges that that trigram's removal would cause. This could result
in a broken final graph that would never match anything, leading to GIN
or GiST indexscans not finding anything.
To fix, add a "tentParent" field that is used only within this loop,
and set it to show state merges that we are tentatively going to do.
While examining a particular arc, we must chase up through tentParent
links as well as regular parent links (the former can only appear atop
the latter), and we must account for state init/fin flag merges that
haven't actually been done yet.
To simplify the latter, combine the separate init and fin bool fields
into a bitmap flags field. I also chose to get rid of the "children"
state list, which seems entirely inessential.
Per bug #14563 from Alexey Isayko, which the added test cases are based on.
Back-patch to 9.3 where this code was added.
Andrew Dunstan [Wed, 22 Feb 2017 16:10:49 +0000 (11:10 -0500)]
Correctly handle array pseudotypes in to_json and to_jsonb
Columns with array pseudotypes have not been identified as arrays, so
they have been rendered as strings in the json and jsonb conversion
routines. This change allows them to be rendered as json arrays, making
it possible to deal correctly with the anyarray columns in pg_stats.
Robert Haas [Wed, 22 Feb 2017 06:45:17 +0000 (12:15 +0530)]
Pass the source text for a parallel query to the workers.
With this change, you can see the query that a parallel worker is
executing in pg_stat_activity, and if the worker crashes you can
see what query it was executing when it crashed.
Rafia Sabih, reviewed by Kuntal Ghosh and Amit Kapila and slightly
revised by me.
Robert Haas [Wed, 22 Feb 2017 02:29:27 +0000 (07:59 +0530)]
Shut down Gather's children before shutting down Gather itself.
It turns out that the original shutdown order here does not work well.
Multiple people attempting to develop further parallel query patches
have discovered that they need to do cleanup before the DSM goes away,
and you can't do that if the parent node gets cleaned up first.
Patch by me, reviewed by KaiGai Kohei and Dilip Kumar.
Tom Lane [Tue, 21 Feb 2017 22:51:27 +0000 (17:51 -0500)]
Fix sloppy handling of corner-case errors in fd.c.
Several places in fd.c had badly-thought-through handling of error returns
from lseek() and close(). The fact that those would seldom fail on valid
FDs is probably the reason we've not noticed this up to now; but if they
did fail, we'd get quite confused.
LruDelete and LruInsert actually just Assert'd that lseek never fails,
which is pretty awful on its face.
In LruDelete, we indeed can't throw an error, because that's likely to get
called during error abort and so throwing an error would probably just lead
to an infinite loop. But by the same token, throwing an error from the
close() right after that was ill-advised, not to mention that it would've
left the LRU state corrupted since we'd already unlinked the VFD from the
list. I also noticed that really, most of the time, we should know the
current seek position and it shouldn't be necessary to do an lseek here at
all. As patched, if we don't have a seek position and an lseek attempt
doesn't give us one, we'll close the file but then subsequent re-open
attempts will fail (except in the somewhat-unlikely case that a
FileSeek(SEEK_SET) call comes between and allows us to re-establish a known
target seek position). This isn't great but it won't result in any state
corruption.
Meanwhile, having an Assert instead of an honest test in LruInsert is
really dangerous: if that lseek failed, a subsequent read or write would
read or write from the start of the file, not where the caller expected,
leading to data corruption.
In both LruDelete and FileClose, if close() fails, just LOG that and mark
the VFD closed anyway. Possibly leaking an FD is preferable to getting
into an infinite loop or corrupting the VFD list. Besides, as far as I can
tell from the POSIX spec, it's unspecified whether or not the file has been
closed, so treating it as still open could be the wrong thing anyhow.
I also fixed a number of other places that were being sloppy about
behaving correctly when the seekPos is unknown.
Also, I changed FileSeek to return -1 with EINVAL for the cases where it
detects a bad offset, rather than throwing a hard elog(ERROR). It seemed
pretty inconsistent that some bad-offset cases would get a failure return
while others got elog(ERROR). It was missing an offset validity check for
the SEEK_CUR case on a closed file, too.
Back-patch to all supported branches, since all this code is fundamentally
identical in all of them.
projects / postgresql / log