| Age | Commit message (Collapse) | Author |
|---|
|
Typos reported by Álvaro Herrera and Erik Rijkers.
Catversion bump for 3d14e171e9e2236139e8976f3309a588bcc8683b was
inadvertently omitted.
|
|
Similar to how the INHERIT option controls whether or not the
permissions of the granted role are automatically available to the
grantee, the new SET permission controls whether or not the grantee
may use the SET ROLE command to assume the privileges of the granted
role.
In addition, the new SET permission controls whether or not it
is possible to transfer ownership of objects to the target role
or to create new objects owned by the target role using commands
such as CREATE DATABASE .. OWNER. We could alternatively have made
this controlled by the INHERIT option, or allow it when either
option is given. An advantage of this approach is that if you
are granted a predefined role with INHERIT TRUE, SET FALSE, you
can't go and create objects owned by that role.
The underlying theory here is that the ability to create objects
as a target role is not a privilege per se, and thus does not
depend on whether you inherit the target role's privileges. However,
it's surely something you could do anyway if you could SET ROLE
to the target role, and thus making it contingent on whether you
have that ability is reasonable.
Design review by Nathan Bossat, Wolfgang Walther, Jeff Davis,
Peter Eisentraut, and Stephen Frost.
Discussion: http://postgr.es/m/CA+Tgmob+zDSRS6JXYrgq0NWdzCXuTNzT5eK54Dn2hhgt17nm8A@mail.gmail.com
|
|
Erik Rijkers
|
|
|
|
Extracted from a more ambitious patch.
Author: David G. Johnston <[email protected]>
Discussion: https://postgr.es/m/CAKFQuwZC4K0XYBm0bwBMDOZySBqhOSekDhLuaw4vPi+ozi8gqQ@mail.gmail.com
|
|
The TAP tests can, to some degree, be controlled by a set of environment
variables. These were however only documented in a README and not in the
main documentation. This adds documentation of these variables, as well
as changes one CPAN reference to a ulink for consistency. While there,
also tag CPAN as an acronym as it's listed in the acronyms section.
Reviewed-by: Michael Paquier <[email protected]>
Discussion: https://postgr.es/m/[email protected]
|
|
The /release/ links are redirected to /dist/ and /pod/release/ to
/release/../view/, so update our links accordingly to avoid 301
redirects.
Discussion: https://postgr.es/m/[email protected]
|
|
|
|
Author: Bharath Rupireddy, Jeff Davis
Discussion: https://postgr.es/m/CALj2ACVTBNA1wfVCsikfhygAbZe6kFY8Oz6PhOyhHyA4vAGouA%40mail.gmail.com
|
|
To improve readability of the TG_ variables definition lists, this moves
the datatypes up to the defined term to avoid having each entry start
with "Data type". This also removes redundant wording that that didn't
carry any information from the descriptions.
Author: Christoph Berg <[email protected]>
Reviewed-by: Dagfinn Ilmari Mannsåker <[email protected]>
Discussion: https://postgr.es/m/[email protected]
|
|
This adds a new psql command \bind that sets query parameters and
causes the next query to be sent using the extended query protocol.
Example:
SELECT $1, $2 \bind 'foo' 'bar' \g
This may be useful for psql scripting, but one of the main purposes is
also to be able to test various aspects of the extended query protocol
from psql and to write tests more easily.
Reviewed-by: Corey Huinker <[email protected]>
Discussion: https://www.postgresql.org/message-id/flat/[email protected]
|
|
Setting archive_library and archive_command at the same time is now an
error. Before, archive_library would take precedence over
archive_command.
Author: Nathan Bossart <[email protected]>
Reviewed-by: Peter Eisentraut <[email protected]>
Reviewed-by: Bharath Rupireddy <[email protected]>
Discussion: https://www.postgresql.org/message-id/20220914222736.GA3042279%40nathanxps13
|
|
We already allow explicitly writing DEFAULT for SET COMPRESSION,
so it seems a bit inflexible and non-orthogonal to not have it
for STORAGE.
Aleksander Alekseev
Discussion: https://postgr.es/m/CAJ7c6TMX9ui+6y3TQFaXJYVpZyBukvqhQbVDJ8OUokeLRhtnpA@mail.gmail.com
|
|
Commit fede15417 introduced FILTER by jamming it into the existing
example introducing HAVING, which seems pedagogically poor to me;
and it added no information about what the keyword actually does.
Not to mention that the claimed output didn't match the sample
data being used in this running example.
Revert that and instead make an independent example using FILTER.
To help drive home the point that it's a per-aggregate filter,
we need to use two aggregates not just one; for consistency
expand all the examples in this segment to do that.
Also adjust the example using WHERE ... LIKE so that it'd produce
nonempty output with this sample data, and show that output.
Back-patch, as the previous patch was. (Sadly, v10 is now out
of scope.)
Discussion: https://postgr.es/m/[email protected]
|
|
Make it clear that the data corresponding to the transactions committed
before confirmed_flush_lsn is not available anymore.
Author: Ashutosh Sharma
Reviewd by: Ashutosh Bapat
Discussion: https://postgr.es/m/CAE9k0P=hiqRXUonnmtS-5Pu8SbO=yF6vcrVBcfEf2+93ng_f5Q@mail.gmail.com
|
|
There are some cases (e.g. when the subscription is created using the
connect = false option) where the remote replication slot was not created
automatically and the user must create it manually before the subscription
can be activated. There was not enough information in the docs for users
to do this easily.
Author: Peter Smith
Reviewd by: Shi yu, Amit Kapila
Discussion: https://postgr.es/m/CAHut+PvqdqOanheWSHDyhQiF+Z-7w=-+k4U+bwbT=b6YQ_hrXQ@mail.gmail.com
|
|
pg_ident_file_mappings.line_number was described as a line number in
pg_ident.conf for a "rule" number, but this should refer to a "map".
The same inconsistent term was used in the main paragraph describing the
view.
Extracted from a patch by the same author. Issue introduced by
a2c8499 where this view has been added.
Author: Julien Rouhaud
Discussion: https://postgr.es/m/20221026031948.cbrnzgy5e7glsq2d@jrouhaud
Backpatch-through: 15
|
|
This gives some additional advice on using row filters and column
lists on publications securely.
Author: Antonin Houska <[email protected]>
Reviewed-by: Euler Taveira <[email protected]>
Discussion: https://www.postgresql.org/message-id/flat/20330.1652105397@antos
|
|
These numbers are strictly-monotone identifiers assigned to each rule
of pg_hba_file_rules and each map of pg_ident_file_mappings when loading
the HBA and ident configuration files, indicating the order in which
they are checked at authentication time, until a match is found.
With only one file loaded currently, this is equivalent to the line
numbers assigned to the entries loaded if one wants to know their order,
but this becomes mandatory once the inclusion of external files is
added to the HBA and ident files to be able to know in which order the
rules and/or maps are applied at authentication. Note that NULL is used
when a HBA or ident entry cannot be parsed or validated, aka when an
error exists, contrary to the line number.
Bump catalog version.
Author: Julien Rouhaud
Discussion: https://postgr.es/m/20220223045959.35ipdsvbxcstrhya@jrouhaud
|
|
This entry was listed as a "string", but it is a "number. The other
fields are correctly described, on a second look.
Reported-by: Nuko Yokohama
Author: Tatsuo Ishii
Discussion: https://postgr.es/m/CAF3Gu1awoVoDP5d0_eN=cR=QkGVwH+OtFvwJkkc5cB_ZMWjyeA@mail.gmail.com
Backpatch-through: 15
|
|
As of this commit, any database or user entry beginning with a slash (/)
is considered as a regular expression. This is particularly useful for
users, as now there is no clean way to match pattern on multiple HBA
lines. For example, a user name mapping with a regular expression needs
first to match with a HBA line, and we would skip the follow-up HBA
entries if the ident regexp does *not* match with what has matched in
the HBA line.
pg_hba.conf is able to handle multiple databases and roles with a
comma-separated list of these, hence individual regular expressions that
include commas need to be double-quoted.
At authentication time, user and database names are now checked in the
following order:
- Arbitrary keywords (like "all", the ones beginning by '+' for
membership check), that we know will never have a regexp. A fancy case
is for physical WAL senders, we *have* to only match "replication" for
the database.
- Regular expression matching.
- Exact match.
The previous logic did the same, but without the regexp step.
We have discussed as well the possibility to support regexp pattern
matching for host names, but these happen to lead to tricky issues based
on what I understand, particularly with host entries that have CIDRs.
This commit relies heavily on the refactoring done in a903971 and
fc579e1, so as the amount of code required to compile and execute
regular expressions is now minimal. When parsing pg_hba.conf, all the
computed regexps needs to explicitely free()'d, same as pg_ident.conf.
Documentation and TAP tests are added to cover this feature, including
cases where the regexps use commas (for clarity in the docs, coverage
for the parsing logic in the tests).
Note that this introduces a breakage with older versions, where a
database or user name beginning with a slash are treated as something to
check for an equal match. Per discussion, we have discarded this as
being much of an issue in practice as it would require a cluster to
have database and/or role names that begin with a slash, as well as HBA
entries using these. Hence, the consistency gained with regexps in
pg_ident.conf is more appealing in the long term.
**This compatibility change should be mentioned in the release notes.**
Author: Bertrand Drouvot
Reviewed-by: Jacob Champion, Tom Lane, Michael Paquier
Discussion: https://postgr.es/m/[email protected]
|
|
56788d215 adjusted the parallel seq scan code so that instead of handing
out a single block at a time to parallel workers, it now hands out ranges
of blocks.
Here we update the documentation which still claimed that workers received
just 1 block at a time.
Reported-by: Zhang Mingli
Discussion: https://postgr.es/m/17c99615-2c3b-4e4e-9d0b-424a66a7bccd@Spark
Backpatch-through: 14, where 56788d215 was added.
|
|
Various test suites use the "openssl" program as part of their setup.
There isn't a way to override which openssl program is to be used,
other than by fiddling with the path, perhaps. This has gotten
increasingly problematic because different versions of openssl have
different capabilities and do different things by default.
This patch checks for an openssl binary in configure and meson setup,
with appropriate ways to override it. This is similar to how "lz4"
and "zstd" are handled, for example. The meson build system actually
already did this, but the result was only used in some places. This
is now applied more uniformly.
Reviewed-by: Michael Paquier <[email protected]>
Discussion: https://www.postgresql.org/message-id/flat/dc638b75-a16a-007d-9e1c-d16ed6cf0ad2%40enterprisedb.com
|
|
As currently designed, with a callback registered in a ERROR_CLEANUP
block, the shutdown callback would get called twice when updating
archive_library on SIGHUP, which is something that we want to avoid to
ease the life of extension writers.
Anyway, an ERROR in the archiver process is treated as a FATAL, stopping
it immediately, hence there is no need for a ERROR_CLEANUP block.
Instead of that, the shutdown callback is not called upon
before_shmem_exit(), giving to the modules the opportunity to do any
cleanup actions before the server shuts down its subsystems.
While on it, this commit adds some testing coverage for the shutdown
callback. Neither shell_archive nor basic_archive have been using it,
and one is added to shell_archive, whose trigger is checked in a TAP
test through a shutdown sequence.
Author: Nathan Bossart, Bharath Rupireddy
Reviewed-by: Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/20221015221328.GB1821022@nathanxps13
Backpatch-through: 15
|
|
It was previously easily overlooked at the end of several tables.
Reported-by: Alex Denman
Discussion: https://postgr.es/m/[email protected]
Backpatch-through: 10
|
|
The fix is to run ANALYZE.
Discussion: https://postgr.es/m/[email protected],
https://postgr.es/m/flat/CAKJS1f8DTbCHf9gedU0He6ARsd58E6qOhEHM1caomqj_r9MOiQ%40mail.gmail.com,
https://postgr.es/m/CAKJS1f80o98hcfSk8j%3DfdN09S7Sjz%2BvuzhEwbyQqvHJb_sZw0g%40mail.gmail.com
Backpatch-through: 10
|
|
It can be useful to know when a relation has last been used, e.g., when
evaluating whether an index is still required. It was already possible to
infer the time of the last usage by tracking, e.g.,
pg_stat_all_indexes.idx_scan over time. But far from everybody does so.
To make it easier to detect the last time a relation has been scanned, track
that time in each relation's pgstat entry. To minimize overhead a) the
timestamp is updated only when the backend pending stats entry is flushed to
shared stats b) the last transaction's stop timestamp is used as the
timestamp.
Bumps catalog and stats format versions.
Author: Dave Page <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Bruce Momjian <[email protected]>
Reviewed-by: Vik Fearing <[email protected]>
Discussion: https://postgr.es/m/CA+OCxozrVHNFVEPkweUHMZje+t1tfY816d9MZYc6eZwOOusOaQ@mail.gmail.com
|
|
This has apparently been wrong since the beginning
(090d0f2050647958865cb495dff74af7257d2bb4).
Discussion: https://www.postgresql.org/message-id/[email protected]
|
|
Using pg_buffercache_summary() is significantly cheaper than querying
pg_buffercache and summarizing in SQL.
Author: Melih Mutlu <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Aleksander Alekseev <[email protected]>
Reviewed-by: Zhang Mingli <[email protected]>
Discussion: https://postgr.es/m/CAGPVpCQAXYo54Q%3D8gqBsS%3Du0uk9qhnnq4%2B710BtUhUisX1XGEg%40mail.gmail.com
|
|
While at it, simplify wording a bit.
Author: Takamichi Osumi <[email protected]>
Reviewed-by: Peter Smith <[email protected]>
Discussion: https://postgr.es/m/TYCPR01MB8373F93F5D094A2BE648990DED259@TYCPR01MB8373.jpnprd01.prod.outlook.com
|
|
Commit 3d956d956 allowed the COPY, but it's done by inserting individual
rows to the foreign table, so it can be inefficient due to the overhead
caused by each round-trip to the foreign server. To improve performance
of the COPY in such a case, this patch allows batch insertion, by
extending the multi-insert machinery in CopyFrom() to the foreign-table
case so that we insert multiple rows to the foreign table at once using
the FDW callback routine added by commit b663a4136. This patch also
allows this for postgres_fdw. It is enabled by the "batch_size" option
added by commit b663a4136, which is disabled by default.
When doing batch insertion, we update progress of the COPY command after
performing the FDW callback routine, to count rows not suppressed by the
FDW as well as a BEFORE ROW INSERT trigger. For consistency, this patch
changes the timing of updating it for plain tables: previously, we
updated it immediately after adding each row to the multi-insert buffer,
but we do so only after writing the rows stored in the buffer out to the
table using table_multi_insert(), which I think would be consistent even
with non-batching mode, because in that mode we update it after writing
each row out to the table using table_tuple_insert().
Andrey Lepikhov, heavily revised by me, with review from Ian Barwick,
Andrey Lepikhov, and Zhihong Yu.
Discussion: https://postgr.es/m/bc489202-9855-7550-d64c-ad2d83c24867%40postgrespro.ru
|
|
The output plugin name is a mandatory option when creating a logical
slot, but the grammar documented was not described as such. While on
it, fix two comments in repl_gram.y to show that TEMPORARY is an
optional grammar choice.
Author: Ayaki Tachikake
Discussion: https://postgr.es/m/OSAPR01MB2852607B2329FFA27834105AF1229@OSAPR01MB2852.jpnprd01.prod.outlook.com
Backpatch-through: 15
|
|
Add
After=network-online.target
Wants=network-online.target
to the suggested unit file for starting a Postgres server.
This delays startup until the network interfaces have been
configured; without that, any attempt to bind to a specific
IP address will fail.
If listen_addresses is set to "localhost" or "*", it might be
possible to get away with the less restrictive "network.target",
but I don't think we need to get into such detail here.
Per suggestion from Pablo Federico.
Discussion: https://postgr.es/m/[email protected]
|
|
Other pg_get_XXXdef() functions are documented, so it seems reasonable
to include this as well.
Ian Barwick
Discussion: https://postgr.es/m/CAB8KJ=hb2QZXdgyrrRjPCw++DsrRcui4fKArWabQ+oij+2x=_w@mail.gmail.com
|
|
Reported-by: Elena Indrupskaya
Discussion: https://postgr.es/m/[email protected]
Author: Elena Indrupskaya
Backpatch-through: 15
|
|
meson/ninja do not support specifying dependencies via globs (as those make it
significantly more expensive to check the current build state). Instead
targets should emit dependency information when running that then can be
cheaply re-checked during future builds.
To handle xmllint and xsltproc invocations in the docs, add and use a wrapper
that uses --load-trace to collect dependency information.
Reviewed-by: Peter Eisentraut <[email protected]>
Author: Nazir Bilal Yavuz <[email protected]>
Author: Andres Freund <[email protected]>
Discussion: https://postgr.es/m/[email protected]
|
|
This reverts commit db0d67db2401eb6238ccc04c6407a4fd4f985832 and
several follow-on fixes. The idea of making a cost-based choice
of the order of the sorting columns is not fundamentally unsound,
but it requires cost information and data statistics that we don't
really have. For example, relying on procost to distinguish the
relative costs of different sort comparators is pretty pointless
so long as most such comparator functions are labeled with cost 1.0.
Moreover, estimating the number of comparisons done by Quicksort
requires more than just an estimate of the number of distinct values
in the input: you also need some idea of the sizes of the larger
groups, if you want an estimate that's good to better than a factor of
three or so. That's data that's often unknown or not very reliable.
Worse, to arrive at estimates of the number of calls made to the
lower-order-column comparison functions, the code needs to make
estimates of the numbers of distinct values of multiple columns,
which are necessarily even less trustworthy than per-column stats.
Even if all the inputs are perfectly reliable, the cost algorithm
as-implemented cannot offer useful information about how to order
sorting columns beyond the point at which the average group size
is estimated to drop to 1.
Close inspection of the code added by db0d67db2 shows that there
are also multiple small bugs. These could have been fixed, but
there's not much point if we don't trust the estimates to be
accurate in-principle.
Finally, the changes in cost_sort's behavior made for very large
changes (often a factor of 2 or so) in the cost estimates for all
sorting operations, not only those for multi-column GROUP BY.
That naturally changes plan choices in many situations, and there's
precious little evidence to show that the changes are for the better.
Given the above doubts about whether the new estimates are really
trustworthy, it's hard to summon much confidence that these changes
are better on the average.
Since we're hard up against the release deadline for v15, let's
revert these changes for now. We can always try again later.
Note: in v15, I left T_PathKeyInfo in place in nodes.h even though
it's unreferenced. Removing it would be an ABI break, and it seems
a bit late in the release cycle for that.
Discussion: https://postgr.es/m/TYAPR01MB586665EB5FB2C3807E893941F5579@TYAPR01MB5866.jpnprd01.prod.outlook.com
|
|
This fixes some areas related to logical replication and custom RMGRs.
Author: Ekaterina Kiryanova
Discussion: https://postgr.es/m/[email protected]
Backpatch-through: 15
|
|
Currently, PQsslAttributeNames() returns the same list of attribute
names regardless of its conn parameter. This patch changes it to
have behavior parallel to what 80a05679d installed for PQsslAttribute:
you get OpenSSL's attributes if conn is NULL or is an SSL-encrypted
connection, or an empty list if conn is a non-encrypted connection.
The point of this is to have sensible connection-dependent behavior
in case we ever support multiple SSL libraries. The behavior for
NULL can be defined as "the attributes for the default SSL library",
parallel to what PQsslAttribute(NULL, "library") does.
Since this is mostly just future-proofing, no back-patch.
Discussion: https://postgr.es/m/[email protected]
|
|
The compression parameter to PQsslAttribute has never returned the
compression method used, it has always returned "on" or "off since
it was added in commit 91fa7b4719ac. Backpatch through v10.
Reviewed-by: Tom Lane <[email protected]>
Discussion: https://postgr.es/m/[email protected]
Backpatch-through: v10
|
|
Commit ebc8b7d44 intended to change the behavior of
PQsslAttribute(NULL, "library"), but accidentally also changed
what happens with a non-NULL conn pointer. Undo that so that
only the intended behavior change happens. Clarify some
associated documentation.
Per bug #17625 from Heath Lord. Back-patch to v15.
Discussion: https://postgr.es/m/[email protected]
|
|
Up to now, the ID values returned by pg_stat_get_backend_idset() and
used by pg_stat_get_backend_activity() and allied functions were just
indexes into a local array of sessions seen by the last stats refresh.
This is problematic for a few reasons. The "ID" of a session can vary
over its existence, which is surprising. Also, while these numbers
often match the "backend ID" used for purposes like temp schema
assignment, that isn't reliably true. We can fairly cheaply switch
things around to make these numbers actually be the sessions' backend
IDs. The added test case illustrates that with this definition, the
temp schema used by a given session can be obtained given its PID.
While here, delete some dead code that guarded against getting
a NULL return from pgstat_fetch_stat_local_beentry(). That can't
happen as long as the caller is careful to pass an in-range array
index, as all the callers are. (This code may not have been dead
when written, but it surely is now.)
Nathan Bossart
Discussion: https://postgr.es/m/20220815205811.GA250990@nathanxps13
|
|
SYSTEM_USER is a reserved keyword of the SQL specification that,
roughly described, is aimed at reporting some information about the
system user who has connected to the database server. It may include
implementation-specific information about the means by the user
connected, like an authentication method.
This commit implements SYSTEM_USER as of auth_method:identity, where
"auth_method" is a keyword about the authentication method used to log
into the server (like peer, md5, scram-sha-256, gss, etc.) and
"identity" is the authentication identity as introduced by 9afffcb (peer
sets authn to the OS user name, gss to the user principal, etc.). This
format has been suggested by Tom Lane.
Note that thanks to d951052, SYSTEM_USER is available to parallel
workers.
Bump catalog version.
Author: Bertrand Drouvot
Reviewed-by: Jacob Champion, Joe Conway, Álvaro Herrera, Michael Paquier
Discussion: https://postgr.es/m/[email protected]
|
|
Reported-by: Tom Lane
Discussion: https://postgr.es/m/[email protected]
Backpatch-through: 10
|
|
doc revert of commit 1703726488. Change was applied to irrelevant
branches, and was not detailed enough to be helpful in relevant
branches.
Reported-by: Peter Eisentraut, Noah Misch
Discussion: https://postgr.es/m/[email protected]
Backpatch-through: 10
|
|
For some reason the "bpchar" type name was defined nowhere in
our SGML docs, although several places refer to it in passing.
Give it a proper mention under Character Types.
While here, also provide an explanation of how the text and varchar
types relate. The previous wording seemed to be doing its best
to sweep text under the rug, which doesn't seem very appropriate
given its prominence in other parts of the docs.
Minor rearrangements and word-smithing for clarity, too.
Laurenz Albe and Tom Lane, per gripe from Yanliang Lei
Discussion: https://postgr.es/m/[email protected]
|
|
There are still some alignment-related failures in the buildfarm,
which might or might not be able to be fixed quickly, but I've also
just realized that it increased the size of many WAL records by 4 bytes
because a block reference contains a RelFileLocator. The effect of that
hasn't been studied or discussed, so revert for now.
|
|
The previous macro implementations just cast the argument to a target
type but did not check whether the input type was appropriate. The
function implementation can do better type checking of the input type.
For the *GetDatumFast() macros, converting to an inline function
doesn't work in the !USE_FLOAT8_BYVAL case, but we can use
AssertVariableIsOfTypeMacro() to get a similar level of type checking.
Reviewed-by: Aleksander Alekseev <[email protected]>
Reviewed-by: Tom Lane <[email protected]>
Discussion: https://www.postgresql.org/message-id/flat/8528fb7e-0aa2-6b54-85fb-0c0886dbd6ed%40enterprisedb.com
|
|
RelFileNumbers are now assigned using a separate counter, instead of
being assigned from the OID counter. This counter never wraps around:
if all 2^56 possible RelFileNumbers are used, an internal error
occurs. As the cluster is limited to 2^64 total bytes of WAL, this
limitation should not cause a problem in practice.
If the counter were 64 bits wide rather than 56 bits wide, we would
need to increase the width of the BufferTag, which might adversely
impact buffer lookup performance. Also, this lets us use bigint for
pg_class.relfilenode and other places where these values are exposed
at the SQL level without worrying about overflow.
This should remove the need to keep "tombstone" files around until
the next checkpoint when relations are removed. We do that to keep
RelFileNumbers from being recycled, but now that won't happen
anyway. However, this patch doesn't actually change anything in
this area; it just makes it possible for a future patch to do so.
Dilip Kumar, based on an idea from Andres Freund, who also reviewed
some earlier versions of the patch. Further review and some
wordsmithing by me. Also reviewed at various points by Ashutosh
Sharma, Vignesh C, Amul Sul, Álvaro Herrera, and Tom Lane.
Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com
|
|
Previously, the transaction-property GUCs such as transaction_isolation
could be reset after starting a transaction, because we marked them
as GUC_NO_RESET_ALL but still allowed a targeted RESET. That leads to
assertion failures or worse, because those properties aren't supposed
to change after we've acquired a transaction snapshot.
There are some NO_RESET_ALL variables for which RESET is okay, so
we can't just redefine the semantics of that flag. Instead introduce
a separate GUC_NO_RESET flag. Mark "seed", as well as the transaction
property GUCs, as GUC_NO_RESET.
We have to disallow GUC_ACTION_SAVE as well as straight RESET, because
otherwise a function having a "SET transaction_isolation" clause can
still break things: the end-of-function restore action is equivalent
to a RESET.
No back-patch, as it's conceivable that someone is doing something
this patch will forbid (like resetting one of these GUCs at transaction
start, or "CREATE FUNCTION ... SET transaction_read_only = 1") and not
running into problems with it today. Given how long we've had this
issue and not noticed, the side effects in non-assert builds can't be
too serious.
Per bug #17385 from Andrew Bille.
Masahiko Sawada
Discussion: https://postgr.es/m/[email protected]
|
