On Sat, Jul 12, 2014 at 1:26 AM, Martijn van Oosterhout
wrote:
> On Thu, Jul 10, 2014 at 08:46:55AM +0530, Amit Kapila wrote:
> > As per my understanding of the general theory around barriers,
> > read and write are defined to avoid reordering due to compiler and
> > full memory barriers are defin
On Fri, Jul 11, 2014 at 11:15 PM, Robert Haas wrote:
> On Thu, Jul 10, 2014 at 1:05 AM, Amit Kapila
wrote:
> > If there is an noticeable impact, then do you think having separate
> > file/infrastructure for parallel sort can help, basically non-parallel
> > and parallel sort will have some common
On Thu, Jun 19, 2014 at 01:01:34PM -0400, Tom Lane wrote:
> Noah Misch writes:
> > On Thu, Jun 12, 2014 at 05:02:19PM -0400, Noah Misch wrote:
> >> You can cause the at-exit crash by building PostgreSQL against OpenLDAP
> >> 2.4.31, connecting with LDAP authentication, and issuing "LOAD 'dblink'".
On Fri, 2014-07-11 at 11:51 -0400, Tom Lane wrote:
> Jeff Davis writes:
> > Attached is a small patch to $SUBJECT.
> > In master, only single-byte characters are allowed as an escape. Of
> > course, with the patch it must still be a single character, but it may
> > be multi-byte.
>
> I'm concerne
On Thu, Jul 10, 2014 at 2:52 AM, Tomonari Katsumata
wrote:
> Several couple weeks ago, I posted a mail to pgsql-doc.
> http://www.postgresql.org/message-id/53992ff8.2060...@po.ntts.co.jp
>
> In this thread, I concluded that it's better to
> round up the value which is less than its unit.
> Current
On Thu, Jul 10, 2014 at 08:46:55AM +0530, Amit Kapila wrote:
> As per my understanding of the general theory around barriers,
> read and write are defined to avoid reordering due to compiler and
> full memory barriers are defined to avoid reordering due to processors.
> There are some processors th
This example in the regression database is a simplified version of a bug
I was shown off-list:
regression=# select (
select q from
( select 1,2,3 where f1>0
union all
select 4,5,6.0 where f1<=0
) q
)
from int4_tbl;
ERROR: record type has not been registered
The reason for the problem is that
On Fri, Jul 11, 2014 at 3:47 PM, Greg Stark wrote:
> On Fri, Jul 11, 2014 at 6:00 PM, Claudio Freire
> wrote:
>> Marking as read-only is ok, or emitting a NOTICE so that if anyone
>> changes those parameters that change the shape of the index, they know
>> it needs a rebuild would be OK too. Bot
On Fri, Jul 11, 2014 at 6:00 PM, Claudio Freire wrote:
> Marking as read-only is ok, or emitting a NOTICE so that if anyone
> changes those parameters that change the shape of the index, they know
> it needs a rebuild would be OK too. Both mechanisms work for me.
We don't actually have any of the
Robert,
On Friday, July 11, 2014, Robert Haas wrote:
> On Fri, Jul 11, 2014 at 4:55 AM, Stephen Frost > wrote:
> > My feeling at the moment is that having them be per-table makes sense and
> > we'd still have flexibility to change later if we had some compelling
> reason
> > to do so.
>
> I don
On Thu, Jul 10, 2014 at 1:05 AM, Amit Kapila wrote:
> On Tue, Jul 8, 2014 at 1:27 AM, Robert Haas wrote:
>> 6. In general, I'm worried that it's going to be hard to keep the
>> overhead of parallel sort from leaking into the non-parallel case.
>> With the no-allocator approach, every place that u
On Fri, Jul 11, 2014 at 4:55 AM, Stephen Frost wrote:
> On Thursday, July 10, 2014, Robert Haas wrote:
>> On Wed, Jul 9, 2014 at 2:13 AM, Stephen Frost wrote:
>> > Yes, this would be possible (and is nearly identical to the original
>> > patch, except that this includes per-role considerations),
Heikki Linnakangas wrote:
> I did again the refactoring you did back in 2006, patch attached.
> One thing I did differently: I moved the raw, non-encrypted,
> read/write functions to separate functions: pqsecure_raw_read and
> pqsecure_raw_write. Those functions encapsulate the SIGPIPE
> handling.
On 10.7.2014 21:33, Tomas Vondra wrote:
> On 9.7.2014 16:07, Robert Haas wrote:
>> On Tue, Jul 8, 2014 at 5:16 PM, Tomas Vondra wrote:
>>> Thinking about this a bit more, do we really need to build the hash
>>> table on the first pass? Why not to do this:
>>>
>>> (1) batching
>>> - read the tu
Sawada Masahiko wrote:
> As you said, if line number reached UINT_MAX then I think that this
> case is too strange.
> I think INT_MAX is enough for line number.
My point is not whether 2 billion is a better number than 4 billion as a
maximum value. My point is that wraparound of signed int is, I
On Fri, Jul 11, 2014 at 11:01 PM, Alvaro Herrera
wrote:
> Jeevan Chalke wrote:
>
>> On Fri, Jul 11, 2014 at 3:13 PM, Sawada Masahiko
>> wrote:
>
>> > And the line number should be switched to 1 when line number has
>> > reached to INT_MAX?
>>
>> Yes, when it goes beyond INT_MAX, wrap around to 1.
On Thu, Jul 10, 2014 at 4:20 PM, Alvaro Herrera
wrote:
> Claudio Freire wrote:
>> On Wed, Jul 9, 2014 at 6:16 PM, Alvaro Herrera
>> wrote:
>> > Another thing I noticed is that version 8 of the patch blindly believed
>> > the "pages_per_range" declared in catalogs. This meant that if somebody
>>
On Fri, Jul 11, 2014 at 12:12 PM, Michael Banck wrote:
> On Fri, Jul 11, 2014 at 11:07:21AM -0400, Tom Lane wrote:
> > David G Johnston writes:
> > > Benedikt Grundmann wrote
> > >> That is it possible to tell the planner that index is off limits
> > >> i.e.
> > >> don't ever generate a plan usi
On Fri, Jul 11, 2014 at 11:07:21AM -0400, Tom Lane wrote:
> David G Johnston writes:
> > Benedikt Grundmann wrote
> >> That is it possible to tell the planner that index is off limits
> >> i.e.
> >> don't ever generate a plan using it?
>
> > Catalog hacking could work but not recommended (nor do
Jeff Davis writes:
> Attached is a small patch to $SUBJECT.
> In master, only single-byte characters are allowed as an escape. Of
> course, with the patch it must still be a single character, but it may
> be multi-byte.
I'm concerned about the performance cost of this patch. Have you done
any me
On Fri, 2014-07-11 at 14:41 +0900, Fujii Masao wrote:
> Could you add the patch into next CF?
Sure. The patch is so small I was thinking about committing it in a few
days (assuming no complaints), but I'm in no hurry.
> The patch doesn't contain the change of the document. But I think that
> it's
Andres Freund writes:
> On 2014-07-11 11:20:08 -0400, Tom Lane wrote:
>> If you're talking about SnapshotNow hazards, I think the risk would be
>> minimal, and probably no worse than cases that the system will cause
>> by itself.
> Yes, SnapshotNow. I could reproduce it causing 'spurious' HOT upd
On 2014-07-11 11:20:08 -0400, Tom Lane wrote:
> Andres Freund writes:
> > On 2014-07-11 11:07:21 -0400, Tom Lane wrote:
> >> Hacking pg_index.indisvalid could work, given a reasonably recent PG.
> >> I would not try it in production until I'd tested it ;-)
>
> > Works, but IIRC can cause problems
Andres Freund writes:
> On 2014-07-11 11:07:21 -0400, Tom Lane wrote:
>> Hacking pg_index.indisvalid could work, given a reasonably recent PG.
>> I would not try it in production until I'd tested it ;-)
> Works, but IIRC can cause problems at least < 9.4 because concurrent
> cache builds might mi
On 2014-07-11 11:07:21 -0400, Tom Lane wrote:
> David G Johnston writes:
> > Benedikt Grundmann wrote
> >> That is it possible to tell the planner that index is off limits i.e.
> >> don't ever generate a plan using it?
>
> > Catalog hacking could work but not recommended (nor do I know the proper
David G Johnston writes:
> Benedikt Grundmann wrote
>> That is it possible to tell the planner that index is off limits i.e.
>> don't ever generate a plan using it?
> Catalog hacking could work but not recommended (nor do I know the proper
> commands and limitations). Do you need the database/ta
David Rowley writes:
> On Fri, Jul 11, 2014 at 1:11 PM, Tom Lane wrote:
>> Hm ... actually, there might be a better answer: what about transforming
>> WHERE (x,y) NOT IN (SELECT provably-not-null-values FROM ...)
>> to
>> WHERE AND x IS NOT NULL AND y IS NOT NULL
> I think this is the way to go
Benedikt Grundmann wrote
> That is it possible to tell the planner that index is off limits i.e.
> don't
> ever generate a plan using it?
>
> Rationale: Schema changes on big tables. I might have convinced myself /
> strong beliefs that for all queries that I need to be fast the planner
> does
>
Jeevan Chalke wrote:
> On Fri, Jul 11, 2014 at 3:13 PM, Sawada Masahiko
> wrote:
> > And the line number should be switched to 1 when line number has
> > reached to INT_MAX?
>
> Yes, when it goes beyond INT_MAX, wrap around to 1.
>
> BTW, I wonder, can't we simply use unsigned int instead?
Th
On Fri, Jul 11, 2014 at 09:48:06AM -0400, Bruce Momjian wrote:
> > Uh, why does this need to be in ALTER TABLE? Can't this be part of
> > table creation done by pg_dump?
>
> Uh, I think you need to read the thread. We have to delay the toast
> creation part so we don't use an oid that will later
On Fri, Jul 11, 2014 at 12:18:40AM -0400, Alvaro Herrera wrote:
> Bruce Momjian wrote:
> > On Thu, Jul 10, 2014 at 06:38:26PM -0400, Bruce Momjian wrote:
>
> > > I have thought some more on this. I thought I would need to open
> > > pg_class in C and do complex backend stuff, but I now realize I
Fujii Masao wrote:
> On Thu, Jul 10, 2014 at 6:16 AM, Alvaro Herrera
> wrote:
> > Here's a new version of this patch, which is more generic the original
> > versions, and similar to what you describe.
>
> I've not read the discussion so far at all, but I found the problem
> when I played with th
Thank you for review.
>So, you're compressing backup blocks one by one. I wonder if that's the
>right idea and if we shouldn't instead compress all of them in one run to
>increase the compression ratio.
The idea behind compressing blocks one by one was to keep the code as much
similar to the origi
On Fri, Jul 11, 2014 at 1:11 PM, Tom Lane wrote:
> I wrote:
> > We could no doubt fix this by also insisting that the left-side vars
> > be provably not null, but that's going to make the patch even slower
> > and even less often applicable. I'm feeling discouraged about whether
> > this is wort
Hi,
On Fri, Jul 11, 2014 at 3:13 PM, Sawada Masahiko
wrote:
> >
> >
>
> To my understating cleanly, you means that line number is not changed
> when newline has reached to INT_MAX, is incorrect?
>
As per my thinking yes.
> And the line number should be switched to 1 when line number has
> re
On Fri, Jul 11, 2014 at 12:01 AM, Vik Fearing wrote:
> On 07/10/2014 09:32 PM, Magnus Hagander wrote:
>> It seems psql is missing autocomplete entries for LC_COLLATE and
>> LC_CTYPE for the CREATE DATABASE command. Attached patch adds that.
>>
>> I doubt this is important enough to backpatch - tho
On 11 July 2014 10:23, Tomas Vondra wrote:
> On 11 Červenec 2014, 9:27, Simon Riggs wrote:
>> On 9 July 2014 18:54, Tomas Vondra wrote:
>>
>>> (1) size the buckets for NTUP_PER_BUCKET=1 (and use whatever number
>>> of batches this requires)
>>
>> If we start off by assuming NTUP_PER_BUCKET =
On Fri, Jul 11, 2014 at 5:45 PM, Etsuro Fujita
wrote:
> I've noticed that EXPLAIN ANALYZE shows no information on exact/lossy
> blocks for a bitmap heap scan when both the numbers of exact/lossy pages
> retrieved by the node are zero. Such an example is shown below. I
> think it would be better
Hi
I noticed a minor inconsistency with the search_path separator used in the
default configuration.
The schemas of any search_path set using `SET search_path TO...` are
separated by ", " (comma, space), while the default value is only separated
by "," (comma).
The attached patch against master
On Wed, Jul 9, 2014 at 12:59 PM, Tom Lane wrote:
> David Rowley writes:
> > On 9 July 2014 09:27, Tom Lane wrote:
> >> On review it looks like analyzejoins.c would possibly benefit from an
> >> earlier fast-path check as well.
>
> > Do you mean for non-subqueries? There already is a check to se
On Fri, Jul 11, 2014 at 4:23 PM, Jeevan Chalke
wrote:
> Hi,
>
> A.
>
> However, this introduced new bug. As I told, when editor number of lines
> reaches INT_MAX it starts giving negative number. You tried overcoming this
> issue by adding "< 0" check. But I guess you again fumbled in setting that
Re: To Bruce Momjian 2014-07-11 <20140711093923.ga3...@msg.df7cb.de>
> Re: Bruce Momjian 2014-07-08 <20140708202114.gd9...@momjian.us>
> > > > > I believe pg_upgrade itself still needs a fix. While it's not a
> > > > > security problem to put the socket in $CWD while upgrading (it is
> > > > > usin
Re: Bruce Momjian 2014-07-08 <20140708202114.gd9...@momjian.us>
> > > > I believe pg_upgrade itself still needs a fix. While it's not a
> > > > security problem to put the socket in $CWD while upgrading (it is
> > > > using -c unix_socket_permissions=0700), this behavior is pretty
> > > > unexpecte
On 2014-07-11 11:18:58 +0200, Magnus Hagander wrote:
> On Fri, Jul 11, 2014 at 11:14 AM, Andres Freund
> wrote:
> > On 2014-07-11 11:08:48 +0200, Magnus Hagander wrote:
> >> Is there a particular reason why pg_receivexlog only supports using
> >> manually created slots but pg_recvlogical supports
On 11 Červenec 2014, 9:27, Simon Riggs wrote:
> On 9 July 2014 18:54, Tomas Vondra wrote:
>
>> (1) size the buckets for NTUP_PER_BUCKET=1 (and use whatever number
>> of batches this requires)
>
> If we start off by assuming NTUP_PER_BUCKET = 1, how much memory does
> it save to recalculate the
I think the following comment for store_returning_result() in
postgres_fdw.c is not right.
/* PGresult must be released before leaving this function. */
I think PGresult should not be released before leaving this function *on
success* in that function.
(I guess the comment has been copied an
On Fri, Jul 11, 2014 at 11:14 AM, Andres Freund wrote:
> On 2014-07-11 11:08:48 +0200, Magnus Hagander wrote:
>> Is there a particular reason why pg_receivexlog only supports using
>> manually created slots but pg_recvlogical supports creating and
>> dropping them? Wouldn't it be good for consiste
On 2014-07-11 11:08:48 +0200, Magnus Hagander wrote:
> Is there a particular reason why pg_receivexlog only supports using
> manually created slots but pg_recvlogical supports creating and
> dropping them? Wouldn't it be good for consistency there?
I've added it to recvlogical because logical deco
Is there a particular reason why pg_receivexlog only supports using
manually created slots but pg_recvlogical supports creating and
dropping them? Wouldn't it be good for consistency there?
I'm guessing it's related to not being exposed over the replication
protocol? We had a discussion earlier th
On Thursday, July 10, 2014, Robert Haas wrote:
> On Wed, Jul 9, 2014 at 2:13 AM, Stephen Frost > wrote:
> > Yes, this would be possible (and is nearly identical to the original
> > patch, except that this includes per-role considerations), however, my
> > thinking is that it'd be simpler to work
I've noticed that EXPLAIN ANALYZE shows no information on exact/lossy
blocks for a bitmap heap scan when both the numbers of exact/lossy pages
retrieved by the node are zero. Such an example is shown below. I
think it would be better to suppress the 'Heap Blocks' line in that
case, based on the s
Greetings,
Attached modified patch to handle xmlCopyNode returning NULL. The patch is
larger because xmlerrcxt must be passed to xml_xmlnodetoxmltype (xmlerrcxt
is needed for calling xml_ereport).
>From libxml2 source, it turns out that the other cases that xmlCopyNode
will return NULL will not h
That is it possible to tell the planner that index is off limits i.e. don't
ever generate a plan using it?
Rationale: Schema changes on big tables. I might have convinced myself /
strong beliefs that for all queries that I need to be fast the planner does
not need to use a given index (e.g. othe
On 9 July 2014 18:54, Tomas Vondra wrote:
> (1) size the buckets for NTUP_PER_BUCKET=1 (and use whatever number
> of batches this requires)
If we start off by assuming NTUP_PER_BUCKET = 1, how much memory does
it save to recalculate the hash bucket at 10 instead?
Resizing sounds like it will
Hi,
Found new issues with latest patch:
> Thank you for reviewing the patch with variable cases.
> I have revised the patch, and attached latest patch.
>
> > A:
> > Will you please explain the idea behind these changes ?
> I thought wrong about adding new to tail of query_buf.
> The latest patch
(2014/07/11 15:50), Etsuro Fujita wrote:
(2014/07/10 18:12), Shigeru Hanada wrote:
IIUC, you mean that tableoid can't be retrieved when a foreign table
is accessed via parent table,
No. What I want to say is that tableoid *can* be retrieved when a
foreign table is accessed via the parent ta
On 10 July 2014 00:13, Alvaro Herrera wrote:
> Josh Berkus wrote:
>> On 07/09/2014 02:16 PM, Alvaro Herrera wrote:
>> > The way it works now, each opclass needs to have three support
>> > procedures; I've called them getOpers, maybeUpdateValues, and compare.
>> > (I realize these names are pretty
On 9 July 2014 23:54, Peter Geoghegan wrote:
> On Wed, Jul 9, 2014 at 2:16 PM, Alvaro Herrera
> wrote:
>> All this being said, I'm sticking to the name "Minmax indexes". There
>> was a poll in pgsql-advocacy
>> http://www.postgresql.org/message-id/53a0b4f8.8080...@agliodbs.com
>> about a new na
58 matches
Mail list logo