Skip to content

P
Postgres FD Implementation
Switch branch/tag
  1. 17 Feb, 2020 4 commits
  3. 16 Feb, 2020 1 commit
    • Tom Lane's avatar
      Try again to work around Windows' ERROR_SHARING_VIOLATION in pg_ctl. · e02ea141
      Tom Lane authored 
      Commit 0da33c76 introduced an unfortunate regression in pg_ctl on
Windows: if the log file specified with -l doesn't exist yet, and
pg_ctl is running with Administrator privileges, then the log file
might get created with permissions that prevent the postmaster from
writing on it.  (It seems that whether this happens depends on whether
the log file is inside the user's home directory or not, and perhaps
on other phase-of-the-moon conditions, which may explain why we failed
to notice it sooner.)

To fix, just don't create the log file if it doesn't exist yet.  The
case where we need to wait obviously only occurs with a pre-existing
log file.

In passing, switch from using fopen() to plain open(), saving a few
cycles.

Per bug #16259 from Jonathan Katz and Heath Lord.  Back-patch to v12,
as the faulty commit was.

Alexander Lakhin

Discussion: https://postgr.es/m/16259-c5ebed32a262a8b1@postgresql.org
      e02ea141
  5. 15 Feb, 2020 5 commits
    • Tom Lane's avatar
      Update obsolete comment. · faade5d4
      Tom Lane authored 
      Noted by Justin Pryzby, though I chose to just rip out the stale text,
as it's in no way relevant to this particular function.

Discussion: https://postgr.es/m/20200212182337.GZ1412@telsasoft.com
      faade5d4
    • Tom Lane's avatar
      Clarify coding in Catalog::AddDefaultValues. · 9d1ec5a8
      Tom Lane authored 
      Make it a bit shorter and better-commented; no functional change.

Alvaro Herrera and Tom Lane

Discussion: https://postgr.es/m/20200212182337.GZ1412@telsasoft.com
      9d1ec5a8
    • Tom Lane's avatar
      Run "make reformat-dat-files". · b78542b9
      Tom Lane authored 
      Mostly to make sure the previous commit didn't break this.

Discussion: https://postgr.es/m/20200212182337.GZ1412@telsasoft.com
      b78542b9
    • Tom Lane's avatar
      Don't require pg_class.dat to contain correct relnatts values. · 86ff085e
      Tom Lane authored 
      Practically everybody who's ever added a column to one of the bootstrap
catalogs has been burnt by the need to update the relnatts field in the
initial pg_class data to match.  Now that we use Perl scripts to
generate postgres.bki, we can have the machines take care of that,
by filling the field during genbki.pl.

While at it, use the BKI_DEFAULTS mechanism to eliminate repetitive
specifications of other column values in pg_class.dat, too.  They
weren't particularly a maintenance problem, but this way is prettier
(certainly the spotty previous usage of BKI_DEFAULTS wasn't pretty).

No catversion bump needed, since this doesn't actually change the
contents of postgres.bki.

Per gripe from Justin Pryzby, though this is quite different from
his originally proposed solution.

Amit Langote, John Naylor, Tom Lane

Discussion: https://postgr.es/m/20200212182337.GZ1412@telsasoft.com
      86ff085e
    • Peter Geoghegan's avatar
      Recreate website's formatting for "website" doc builds. · 317906f2
      Peter Geoghegan authored 
      The stylesheets used for the HTML documentation rendered on
postgresql.org have shifted, and no longer matched what was expected by
"make STYLE=website html" builds performed locally.  Local doc builds
did not reflect other aspects of the website, including font and
margins.

This patch updates the references to use the current set of stylesheets
that are used by the documentation on postgresql.org. This also wraps
the documentation preview in a HTML container so it can keep the content
within similar margins to those found on the website.

The documentation on building the docs is updated to reflect this
change, and to let the documentation builder know that an external
network connection is required to properly preview documentation built
with "make STYLE=website html" (which was true prior to this patch too,
but not mentioned).

Author: Jonathan Katz
Reported-By: Tom Lane
Discussion: https://postgr.es/m/1375.1581446233@sss.pgh.pa.us
      317906f2
  7. 14 Feb, 2020 2 commits
  9. 13 Feb, 2020 3 commits
    • Tom Lane's avatar
      Mark some contrib modules as "trusted". · eb67623c
      Tom Lane authored 
      This allows these modules to be installed into a database without
superuser privileges (assuming that the DBA or sysadmin has installed
the module's files in the expected place).  You only need CREATE
privilege on the current database, which by default would be
available to the database owner.

The following modules are marked trusted:

btree_gin
btree_gist
citext
cube
dict_int
earthdistance
fuzzystrmatch
hstore
hstore_plperl
intarray
isn
jsonb_plperl
lo
ltree
pg_trgm
pgcrypto
seg
tablefunc
tcn
tsm_system_rows
tsm_system_time
unaccent
uuid-ossp

In the future we might mark some more modules trusted, but there
seems to be no debate about these, and on the whole it seems wise
to be conservative with use of this feature to start out with.

Discussion: https://postgr.es/m/32315.1580326876@sss.pgh.pa.us
      eb67623c
    • Jeff Davis's avatar
      Logical Tape Set: lazily allocate read buffer. · 7fdd919a
      Jeff Davis authored 
      The write buffer was already lazily-allocated, so this is more
symmetric. It also means that a freshly-rewound tape (whether for
reading or writing) is not consuming memory for the buffer.

Discussion: https://postgr.es/m/97c46a59c27f3c38e486ca170fcbc618d97ab049.camel%40j-davis.com
      7fdd919a
    • Tom Lane's avatar
      Avoid a performance regression in float overflow/underflow detection. · 607f8ce7
      Tom Lane authored 
      Commit 6bf0bc84 replaced float.c's CHECKFLOATVAL() macro with static
inline subroutines, but that wasn't too well thought out.  In the original
coding, the unlikely condition (isinf(result) or result == 0) was checked
first, and the inf_is_valid or zero_is_valid condition only afterwards.
The inline-subroutine coding caused that to be swapped around, which is
pretty horrid for performance because (a) in common cases the is_valid
condition is twice as expensive to evaluate (e.g., requiring two isinf()
calls not one) and (b) in common cases the is_valid condition is false,
requiring us to perform the unlikely-condition check anyway.  Net result
is that one isinf() call becomes two or three, resulting in visible
performance loss as reported by Keisuke Kuroda.

The original fix proposal was to revert the replacement of the macro,
but on second thought, that macro was just a bad idea from the beginning:
if anything it's a net negative for readability of the code.  So instead,
let's just open-code all the overflow/underflow tests, being careful to
test the unlikely condition first (and mark it unlikely() to help the
compiler get the point).

Also, rather than having N copies of the actual ereport() calls, collapse
those into out-of-line error subroutines to save some code space.  This
does mean that the error file/line numbers won't be very helpful for
figuring out where the issue really is --- but we'd already burned that
bridge by putting the ereports into static inlines.

In HEAD, check_float[48]_val() are gone altogether.  In v12, leave them
present in float.h but unused in the core code, just in case some
extension is depending on them.

Emre Hasegeli, with some kibitzing from me and Andres Freund

Discussion: https://postgr.es/m/CANDwggLe1Gc1OrRqvPfGE=kM9K0FSfia0hbeFCEmwabhLz95AA@mail.gmail.com
      607f8ce7
  11. 12 Feb, 2020 7 commits
    • Peter Geoghegan's avatar
      Doc: Restructure B-Tree support routine docs. · caba0910
      Peter Geoghegan authored 
      Use a top-level "variablelist", with one item per B-Tree support
function.  This structure matches the structure used by various
"Extensibility" sections in other documentation chapters for other index
access methods.

An explicit list makes it much clearer where each item begins and ends.
This wasn't really a problem before now, but an upcoming patch that adds
deduplication to nbtree will need to have its own new B-Tree support
function.  Ease the burden of translators by tidying up btree.sgml ahead
of committing the deduplication patch.
      caba0910
    • Tom Lane's avatar
      Remove long-dead comments. · 0973f560
      Tom Lane authored 
      These should've been dropped by a8bb8eb5, but evidently were
missed.  Not important enough to back-patch.
      0973f560
    • Tom Lane's avatar
      Doc: fix old oversights in GRANT/REVOKE documentation. · dce98814
      Tom Lane authored 
      The GRANTED BY clause in GRANT/REVOKE ROLE has been there since 2005
but was never documented.  I'm not sure now whether that was just an
oversight or was intentional (given the limited capability of the
option).  But seeing that pg_dumpall does emit code that uses this
option, it seems like not documenting it at all is a bad idea.

Also, when we upgraded the syntax to allow CURRENT_USER/SESSION_USER
as the privilege recipient, the role form of GRANT was incorrectly
not modified to show that, and REVOKE's docs weren't touched at all.

Although I'm not that excited about GRANTED BY, the other oversight
seems serious enough to justify a back-patch.

Discussion: https://postgr.es/m/3070.1581526786@sss.pgh.pa.us
      dce98814
    • Andres Freund's avatar
      Try to harden insert-conflict-specconflict against autovacuum. · 997563df
      Andres Freund authored 
      Looks like guaibasaurus had a autovacuum running during the
controller_print_speculative_locks step (just added in
43e08419). Which does indeed seem quite possible.

Avoid the problem by only looking for the backends participating in
the test.
      997563df
    • Michael Paquier's avatar
      Add %x to default PROMPT1 and PROMPT2 in psql · dcdbb5a5
      Michael Paquier authored 
      %d can be used to track if the current connection is in a transaction
block or not, and adding it by default to the prompt has the advantage
to not need a modification of .psqlrc, something not possible depending
on the environment.

This discussion has happened across various sources, and there was a
strong consensus in favor of this change.

Author: Vik Fearing
Reviewed-by: Fabien Coelho
Discussion: https://postgr.es/m/09502c40-cfe1-bb29-10f9-4b3fa7b2bbb2@2ndquadrant.com
      dcdbb5a5
    • Andres Freund's avatar
      Test additional speculative conflict scenarios. · 43e08419
      Andres Freund authored 
      Previously, the speculative insert tests did not cover the case when a
tuple t is inserted into a table with a unique index on a column but
before it can insert into the index, a concurrent transaction has
inserted a conflicting value into the index and the insertion of tuple t
must be aborted.

The basic permutation is one session successfully inserts into the table
and an associated unique index while a concurrent session successfully
inserts into the table but discovers a conflict before inserting into
the index and must abort the insertion.

Several variants on this include:
- swap which session is successful
- first session insert transaction does not commit, so second session
must wait on a transaction lock
- first session insert does not "complete", so second session must wait
on a speculative insertion lock

Also, refactor the existing TOAST table upsert test to be in the same
spec and reuse the steps.

Author: Melanie Plageman, Ashwin Agrawal, Andres Freund
Reviewed-by: Andres Freund, Taylor Vesely
Discussion: https://postgr.es/m/CAAKRu_ZRmxy_OEryfY3G8Zp01ouhgw59_-_Cm8n7LzRH5BAvng@mail.gmail.com
      43e08419
    • Fujii Masao's avatar
      Fix bug in pg_basebackup -F plain -R. · be6221e9
      Fujii Masao authored 
      Commit caba97a9 changed pg_basebackup -F plain -R so that
it overwrote postgresql.auto.conf in the backup, with new connection
setting. This could cause the existing postgresql.auto.conf settings
in the server to get lost unexpectedly. This is a bug.

This commit fixes the bug by making pg_basebackup -F plain -R
append the connection setting into postgresql.auto.conf in the backup.

Author: Fujii Masao
Reviewed-by: Sergei Kornilov
Discussion: https://postgr.es/m/250dcf2a-94e7-c05e-824a-73cfb38a48a4@oss.nttdata.com
      be6221e9
  13. 11 Feb, 2020 2 commits
  15. 10 Feb, 2020 11 commits
  17. 09 Feb, 2020 2 commits
  19. 07 Feb, 2020 3 commits
    • Alvaro Herrera's avatar
      Fix failure to create FKs correctly in partitions · 55173d2e
      Alvaro Herrera authored 
      On a multi-level partioned table, when adding a partition not directly
connected to the root table, foreign key constraints referencing the
root were not cloned to the new partition, leading to the FK being
possibly inadvertently violated later on.

This was caused by fuzzy thinking in CloneFkReferenced (commit
f56f8f8d): it was skipping constraints marked as having parents on
the theory that cloning those would create duplicates; but that's only
correct for the top level of the partitioning hierarchy.  For levels
below that one, such constraints must still be considered and only
skipped if later on we see that we'd create duplicates.  Apparently, I
(Álvaro) wrote the comments right but the code implemented something
slightly different.

Author: Jehan-Guillaume de Rorthais
Discussion: https://postgr.es/m/20200206004948.238352db@firost
      55173d2e
    • Alvaro Herrera's avatar
      Fix TRUNCATE .. CASCADE on partitions · 9710d3d4
      Alvaro Herrera authored 
      When running TRUNCATE CASCADE on a child of a partitioned table
referenced by another partitioned table, the truncate was not applied to
partitions of the referencing table; this could leave rows violating the
constraint in the referencing partitioned table.  Repair by walking the
pg_constraint chain all the way up to the topmost referencing table.

Note: any partitioned tables containing FKs that reference other
partitioned tables should be checked for possible violating rows, if
TRUNCATE has occurred in partitions of the referenced table.

Reported-by: Christophe Courtois
Author: Jehan-Guillaume de Rorthais
Discussion: https://postgr.es/m/20200204183906.115f693e@firost
      9710d3d4
    • Fujii Masao's avatar
      Fix bug in Tid scan. · cb5b2861
      Fujii Masao authored 
      Commit 147e3722 changed Tid scan so that it calls table_beginscan()
and uses the scan option for seq scan. This change caused two issues.

(1) The change caused Tid scan to take a predicate lock on the entire
       relation in serializable transaction even when relation-level
       lock is not necessary. This could lead to an unexpected
       serialization error.

(2) The change caused Tid scan to increment the number of seq_scan
       in pg_stat_*_tables views even though it's not seq scan. This
       could confuse the users.

This commit adds the scan option for Tid scan and makes Tid scan
use it, to avoid those issues.

Back-patch to v12, where the bug was introduced.

Author: Tatsuhito Kasahara
Reviewed-by: Kyotaro Horiguchi, Masahiko Sawada, Fujii Masao
Discussion: https://postgr.es/m/CAP0=ZVKy+gTbFmB6X_UW0pP3WaeJ-fkUWHoD-pExS=at3CY76g@mail.gmail.com
      cb5b2861