On 2/13/22 01:28, Andres Freund wrote:
> Hi,
>
> On February 12, 2022 3:57:28 PM PST, Thomas Munro
> wrote:
>> Hi,
>>
>> I don't know what happened here, but I my little script that scrapes
>> for build farm assertion failures hasn't seen this one before now.
>> It's on a 15 year old macOS
On 2/12/22 20:58, Tomas Vondra wrote:
> On 2/12/22 01:34, Tomas Vondra wrote:
>> On 2/10/22 19:17, Tomas Vondra wrote:
>>> I've polished & pushed the first part adding sequence decoding
>>> infrastructure etc. Attached are the two remaining parts.
>>>
>>> I plan to wait a day or two and then
Hi,
On 2022-02-12 23:19:38 -0600, Justin Pryzby wrote:
> On Sat, Feb 12, 2022 at 05:20:08PM -0800, Andres Freund wrote:
> > What was the reason behind moving the docs stuff from the compiler warnings
> > task to linux?
>
> I wanted to build docs even if the linux task fails. To allow CFBOT to
Re: Robert Haas
> pg_upgrade: Preserve relfilenodes and tablespace OIDs.
> src/bin/pg_dump/pg_dumpall.c | 3 +
--- a/src/bin/pg_dump/pg_dumpall.c
+++ b/src/bin/pg_dump/pg_dumpall.c
@@ -1066,6 +1066,9 @@ dumpTablespaces(PGconn *conn)
/* needed for
On Thu, Feb 10, 2022 at 3:15 PM Ashutosh Sharma wrote:
>
> On Wed, Feb 9, 2022 at 11:53 PM Euler Taveira wrote:
> >
> > On Wed, Feb 9, 2022, at 12:06 PM, Ashutosh Sharma wrote:
> >
> > Just wondering if we should also be detecting the incorrect conninfo
> > set with ALTER SUBSCRIPTION command as
В Пн, 07/02/2022 в 13:52 +0530, Dilip Kumar пишет:
> On Mon, Jan 31, 2022 at 11:50 AM Kyotaro Horiguchi
> wrote:
> > At Mon, 17 Jan 2022 09:27:14 +0530, Dilip Kumar
> > wrote in
> >
> > me> Mmm. The size of the array cannot be larger than the numbers the
> > me> *Connt() functions return.
> On Wed, Feb 09, 2022 at 01:49:30PM -0500, Melanie Plageman wrote:
> Hi,
> v5 attached and all email feedback addressed below
Thanks for the patch, it looks quite good.
I don't see it in the discussion, so naturally curious -- why directmgr
is not used for bloom index, e.g. in blbuildempty?
>
On Sat, Feb 12, 2022 at 10:51 PM Andres Freund wrote:
> Any reason for using int return types?
>
> particularly since the pg_*_overflow stuff uses bool?
I chose int return types to keep all these methods
consistent with DecodeInterval, which returns a
non-zero int to indicate an error. Though I
On Wed, Jan 26, 2022 at 10:10 PM Peter Eisentraut
wrote:
> [v8 patch]
0001-0004 seem pretty straightforward.
0005:
{realfail1} {
- /*
- * throw back the [Ee], and figure out whether what
- * remains is an {integer} or {decimal}.
- */
- yyless(yyleng - 1);
SET_YYLLOC();
- return
Hi,
In the code comments, it is being used as "Startup process" in the
middle of the sentences whereas in most of the places "startup
process" is used. Also, "WAL Sender" is being used instead of "WAL
sender". Let's be consistent across the docs and code comments.
Attaching a patch. Thoughts?
On Sun, Feb 13, 2022 at 3:13 PM Bharath Rupireddy
wrote:
>
> Hi,
>
> In the code comments, it is being used as "Startup process" in the
> middle of the sentences whereas in most of the places "startup
> process" is used. Also, "WAL Sender" is being used instead of "WAL
> sender". Let's be
On 2/13/22 13:27, Tomas Vondra wrote:
>
> ...
>
> BTW it's probably just a coincidence, but the cfbot does show a strange
> failure on the macOS machine [1], as if we did not wait for all changes
> to be applied by replica. I thought it's because of the issue with
> tracking LSN for sequences
On Fri, Feb 11, 2022 at 7:14 PM Ashutosh Sharma wrote:
>
> I have spent little time looking at the latest patch. The patch looks
> to be in good shape as it has already been reviewed by many people
> here, although I did get some comments. Please take a look and let me
> know your thoughts.
>
>
>
I wrote:
> Justin Pryzby writes:
>> Is there any check for warnings from new code, other than those buildfarm
>> members with -Werror ?
> I periodically grep the buildfarm logs for interesting warnings.
> There are a lot of uninteresting ones :-(, so I'm not sure how
> automatable that'd be. I
Actually found an additional overflow issue in
DecodeInterval. I've updated the patch with the
fix. Specifically it prevents trying to negate a field
if it equals INT_MIN. Let me know if this is best
put into another patch.
- Joe Koshakow
From 09eafa9b496c8461f2dc52ea62c9e833ab10a17f Mon Sep 17
Em ter., 25 de ago. de 2020 às 12:42, Ranier Vilela
escreveu:
> Hi Tom,
>
> Per Coverity.
>
> The SearchSysCache1 allows return NULL and at function AlterStatistics,
> has one mistake, lack of, check of return, which enables a dereference
> NULL pointer,
> at function heap_modify_tuple.
>
>
On Sat, Feb 12, 2022 at 02:26:25PM -0800, Andres Freund wrote:
> On 2022-02-12 16:06:40 -0600, Justin Pryzby wrote:
> > I had some success with that, but it doesn't seem to be significantly
> > faster -
> > it looks a lot like the tests are not actually running in parallel.
Note that the total
Em dom., 13 de fev. de 2022 às 18:43, Andres Freund
escreveu:
> Hi,
>
> On 2022-02-13 18:21:38 -0300, Ranier Vilela wrote:
> > Why open and lock the table Extended Statistics if it is not the owner.
> > Check and return to avoid this.
>
> I was about to say that this opens up
On 2022-02-14 11:23:18 +1300, Thomas Munro wrote:
> On Sun, Feb 13, 2022 at 6:29 PM Andres Freund wrote:
> > Afaics all the "regress test inside tap test" cases would need to do is to
> > pass
> > --outputdir=${PostgreSQL::Test::Utils::tmp_check} and you'd get exactly the
> > same path as
> >
I wrote:
> This conversation motivated me to take a fresh pass over said filtering
> script, and one thing I notice that it's been ignoring is these complaints
> from all the AIX animals:
> ld: 0711-224 WARNING: Duplicate symbol: .xml_is_well_formed
> ld: 0711-224 WARNING: Duplicate symbol:
Hi,
On 2022-02-13 12:16:18 -0800, Andres Freund wrote:
> Waiting for docbook to fix this seems a bit futile, I eventually found a
> bugreport about this, from 2016:
> https://sourceforge.net/p/docbook/bugs/1384/
Now also reported to the current repo:
Chapman Flack writes:
> On 02/13/22 15:16, Tom Lane wrote:
>> Why not just provide equivalents to GetConfigOption() that can
>> deliver int, float8, etc values instead of strings?
> And repeat the bsearch to find the option every time the interested
> extension wants to check the value, even
Hi,
On 2022-02-13 15:36:16 -0500, Chapman Flack wrote:
> Clearly I'm not thinking here of the GUCs that are read-only views of
> values that are determined some other way. How many of those are there
> that are mutable, and could have their values changed without going
> through the GUC
Hi,
Commit
https://github.com/postgres/postgres/commit/6d554e3fcd6fb8be2dbcbd3521e2947ed7a552cb
has fixed the bug, but still has room for improvement.
Why open and lock the table Extended Statistics if it is not the owner.
Check and return to avoid this.
trivial patch attached.
regards,
Hi,
On 2022-02-13 15:31:20 -0600, Justin Pryzby wrote:
> Oh - I suppose you're right. That's an unfortunate consequence of running a
> single prove instance without chdir.
I don't think it's chdir that's relevant (that changes into the source
directory after all). It's the TESTDIR environment
On 02/13/22 15:48, Tom Lane wrote:
> much fix any worries there. Or we could provide APIs that let an
> extension look up a "struct config_generic *" once and then fetch
> directly using that pointer. (But I'd prefer not to, since it'd
> constrain the internals more than I think is wise.)
There
On Sun, Feb 13, 2022 at 1:34 AM Dilip Kumar wrot>
> test4:
> 32 GB shared buffers, template DB size = 10GB, dirty shared buffer=70%
> Head: 47656 ms
> Patch: 79767 ms
This seems like the most surprising result of the bunch. Here, the
template DB is both small enough to fit in shared_buffers and
Hi,
On 2022-02-13 13:27:08 +0100, Tomas Vondra wrote:
> I'm not sure how could this be related to the issue fixed by b779d7d8fd.
> That's merely about handling of empty xacts in the output plugin, it has
> little (nothing) to do with reorderbuffer - the failures was simply
> about (not) printing
Hi,
Sometime last year I was surprised to see (not on a public list unfortunately)
that bookindex.html is 657kB, with > 200kB just being repetitions of
xmlns="http://www.w3.org/1999/xhtml; xmlns:xlink="http://www.w3.org/1999/xlink;
Reminded of this, due to a proposal to automatically generate
On 02/13/22 15:16, Tom Lane wrote:
> That seems like about 10X more complexity than is warranted,
> not only in terms of the infrastructure required, but also in
> the intellectual complexity around "just when could that value
> change?"
>
> Why not just provide equivalents to GetConfigOption()
Hi,
On 2022-02-13 15:02:50 -0600, Justin Pryzby wrote:
> On Sat, Feb 12, 2022 at 04:24:20PM -0800, Andres Freund wrote:
> > > What I am excited about is that some of your other changes showed that we
> > > don't need separate *_artifacts for separate directories anymore. That
> > > used to
> > >
My recent commit 44fa8488 made vacuumlazy.c always scan the last page
of every heap relation, even when it's all-visible, regardless of any
other factor. The general idea is to always examine the last page in
the relation when that might prevent us from unsuccessfully attempting
to truncate the
Hi,
On 2022-02-13 18:21:38 -0300, Ranier Vilela wrote:
> Why open and lock the table Extended Statistics if it is not the owner.
> Check and return to avoid this.
I was about to say that this opens up time-to-check-time-to-use type
issues. But it's the wrong thing to lock to prevent those.
Tomas Vondra writes:
> On 2/13/22 22:43, Andres Freund wrote:
>> Having looked briefly at it, I don't understand what the locking scheme is?
> I don't recall the details of the discussion (if at all), but if you try
> to do concurrent ALTER STATISTICS, that'll end up with:
> ERROR: tuple
Hi,
On 2022-02-13 15:09:20 -0500, Tom Lane wrote:
> Andres Freund writes:
> > On 2022-02-13 12:13:17 -0500, Tom Lane wrote:
> >> Right. Can we set things up so that it's not too painful to inject
> >> custom build options into a CI build?
>
> > What kind of injection are you thinking about?
>
Hi,
On 2022-02-13 15:42:13 -0600, Justin Pryzby wrote:
> > Note that prove unfortunately serializes the test output to be in the order
> > it
> > started them, even when tests run concurrently. Extremely unhelpful, but ...
>
> Are you sure ?
Somewhat. I think it's a question of the prove
Hi,
A recent cfbot failure reminded me of a topic I wanted to raise: The use of
BAIL in tap tests.
When a tap test exits with BAIL it doesn't just consider that test failed, but
it also stops running further tap tests in the same group:
https://metacpan.org/pod/Test::More
"Indicates to the
Andres Freund writes:
> On 2022-02-13 12:13:17 -0500, Tom Lane wrote:
>> Right. Can we set things up so that it's not too painful to inject
>> custom build options into a CI build?
> What kind of injection are you thinking about?
That's exactly what needs to be decided.
> A patch author can
Hi,
On 2022-02-13 09:35:47 -0500, Joseph Koshakow wrote:
> I chose int return types to keep all these methods
> consistent with DecodeInterval, which returns a
> non-zero int to indicate an error.
That's different, because it actually returns different types of errors. IMO
that difference is
(just catching up on this thread)
On Thu, 13 Jan 2022 at 07:20, Robert Haas wrote:
> Yeah. I don't think it's only non-core code we need to worry about
> either. What if I just do EXPLAIN ANALYZE on a prepared query that
> ends up pruning away some stuff? IIRC, the pruned subplans are not
>
On 2/13/22 22:43, Andres Freund wrote:
> Hi,
>
> On 2022-02-13 18:21:38 -0300, Ranier Vilela wrote:
>> Why open and lock the table Extended Statistics if it is not the owner.
>> Check and return to avoid this.
>
> I was about to say that this opens up time-to-check-time-to-use type
> issues. But
On Sun, Feb 13, 2022 at 3:30 AM Andres Freund wrote:
> > There's other things that it'd be nice to enable, but maybe these don't
> > need to
> > be on by default. Maybe just have a list of optional compiler flags (and
> > hints
> > when they're useful). Like WRITE_READ_PARSE_PLAN_TREES.
>
> I
Hi,
On 2022-02-13 12:13:17 -0500, Tom Lane wrote:
> Robert Haas writes:
> > This is exactly why I'm not a huge fan of having ci stuff in the tree.
> > It supposes that there's one right way to do a build, but in reality,
> > different people want and indeed need to use different options for all
Em sex., 11 de jun. de 2021 às 23:05, Ranier Vilela
escreveu:
> Hi,
>
> Removing legitimate warnings can it be worth it?
>
> -1 CAST can be wrong, when there is an invalid value defined
> (InvalidBucket, InvalidBlockNumber).
> I think depending on the compiler -1 CAST may be different from
>
Andres Freund writes:
> On 2022-02-13 15:36:16 -0500, Chapman Flack wrote:
>> Also, I think there are some options that are only represented by
>> an int, float8, etc., when shown, but whose native internal form
>> is something else, like a struct. I was definitely contemplating
>> that you could
Attached is a new version switching from ints to bools, as requested.
- Joe Koshakow
From af8f030ad146602b4386f77b5664c6013743569b Mon Sep 17 00:00:00 2001
From: Joseph Koshakow
Date: Fri, 11 Feb 2022 14:18:32 -0500
Subject: [PATCH] Check for overflow when decoding an interval
When decoding an
On Sat, Feb 12, 2022 at 11:12 PM Tom Lane wrote:
> Eh? copperhead for one is failing for exactly this reason:
>
> Bailout called. Further testing stopped: failed to execute command "lz4 -d
> -m
>
Robert Haas writes:
> This is exactly why I'm not a huge fan of having ci stuff in the tree.
> It supposes that there's one right way to do a build, but in reality,
> different people want and indeed need to use different options for all
> kinds of reasons. That's the whole value of having things
Hi,
On 02/13/22 02:29, Julien Rouhaud wrote:
> Maybe we could have an actually usable GUC API to retrieve values in their
> native format rather than C string for instance, that we could make sure also
> works for cases like max_backend?
I proposed a sketch of such an API for discussion back in
I mentioned this issue briefly in another thread, but the
justify_interval, justify_days, and justify_hours functions
all have the potential to overflow. The attached patch fixes
this issue.
Cheers,
Joe Koshakow
From 4400f2e6886097e3b75d455aeec1ffa9cbc88510 Mon Sep 17 00:00:00 2001
From: Joseph
Andres Freund writes:
> Do we want to consider backpatching these fixes?
I wasn't thinking that we should. It's a behavioral change and
people might not be pleased to have it appear in minor releases.
regards, tom lane
On Sat, Feb 12, 2022 at 04:24:20PM -0800, Andres Freund wrote:
> > What I am excited about is that some of your other changes showed that we
> > don't need separate *_artifacts for separate directories anymore. That used
> > to
> > be the case, but an array of paths is now supported. Putting log,
On Sun, Feb 13, 2022 at 01:23:16PM -0800, Andres Freund wrote:
> If you're seeing this on windows on one of your test branches, that's much
> more likely to be caused by the alltaptests stuff, than by the change in
> artifact instruction.
Oh - I suppose you're right. That's an unfortunate
On Sun, Feb 13, 2022 at 6:29 PM Andres Freund wrote:
> Afaics all the "regress test inside tap test" cases would need to do is to
> pass
> --outputdir=${PostgreSQL::Test::Utils::tmp_check} and you'd get exactly the
> same path as
> REGRESS_OUTPUTDIR currently provides.
Ahh, right. I assume it
Chapman Flack writes:
> On 02/13/22 02:29, Julien Rouhaud wrote:
>> Maybe we could have an actually usable GUC API to retrieve values in their
>> native format rather than C string for instance, that we could make sure also
>> works for cases like max_backend?
> I proposed a sketch of such an
Andres Freund writes:
> Does anybody want to defend / explain the use of BAIL_OUT? If not, I think we
> should consider doing a global replace of the use of bailing.
+1
> Best with a
> central function signalling fatal error, rather than individual uses of die
> or such.
Huh, doesn't
On 2/13/22 13:59, Tom Lane wrote:
> So we're faced with a dilemma: we can't rename either instance without
> breaking something. The only way to get rid of the warnings seems to be
> to decide that the copy in xml2 has outlived its usefulness and remove
> it. I don't think that's a hard
On 2/12/22 16:42, Tom Lane wrote:
>> I'm of the impression that some people have sql access to BF logs.
> Yeah. I'm not sure exactly what the access policy is for that machine;
> I know we'll give out logins to committers, but not whether there's
> any precedent for non-committers.
Yeah, I
Hi,
On 2022-02-13 18:32:59 -0500, Tom Lane wrote:
> Andres Freund writes:
> > Best with a
> > central function signalling fatal error, rather than individual uses of die
> > or such.
>
> Huh, doesn't Test::More already provide a sane way to do this?
I looked, and didn't see anything. But I'm
On Sat, Feb 12, 2022 at 11:12:50PM -0500, Tom Lane wrote:
> Michael Paquier writes:
>> Well, this somebody is I, and the buildfarm did not blow up on any of
>> that so that looked rather fine.
>
> Eh? copperhead for one is failing for exactly this reason:
>
> Bailout called. Further testing
On Sun, Feb 13, 2022 at 11:13:51AM -0500, Robert Haas wrote:
> Well, that's because I didn't realize that LZ4 might be set to
> something that doesn't work at all. So Michael's thing worked, but
> because it (in my view) fixed the problem in a more surprising place
> than I would have preferred, I
On Sun, Feb 13, 2022 at 12:04 PM Dilip Kumar wrote:
>
> On Sun, Feb 13, 2022 at 10:12 AM Dilip Kumar wrote:
> Next, I am planning to do some more tests, where we are having pgbench
> running and concurrently we do CREATEDB maybe every 1 minute and see
> what is the CREATEDB time as well as what
On Sun, Feb 13, 2022 at 02:55:26PM -0800, Andres Freund wrote:
> On 2022-02-14 11:23:18 +1300, Thomas Munro wrote:
>> Ahh, right. I assume it still needs perl2host() treatment for MSYS2
>> systems, because jacana's log shows TESTDIR is set to a Unixoid path
>> that I assume pg_regress's runtime
On Sat, Feb 12, 2022 at 05:31:55PM +0100, Christoph Berg wrote:
> To support multiple major versions in parallel, we need
> /usr/lib/postgresql/NN/lib and /usr/share/postgresql/NN, so it's not a
> single $PREFIX. But you are correct, and ./configure supports that.
Yep, I don't quite see the
Thomas Munro writes:
> We see a successful compile and then a failure to read the file while
> linking. We see that the animal got into that state recently and then
> fixed itself, and now it's back in that state. I don't know if it's
> significant, but it happened to fix itself when a
On Fri, Feb 11, 2022 at 07:50:44AM -0800, Andres Freund wrote:
> On 2022-02-11 16:33:11 +0100, Juan José Santamaría Flecha wrote:
>> The problem comes from the file descriptor passed to gzdopen() in
>> 'src/bin/pg_basebackup/walmethods.c'. Using gzopen() instead, solves the
>> issue without
On Thu, Feb 10, 2022 at 2:35 AM Tom Lane wrote:
> Alexander Korotkov writes:
> > I've rechecked that the now patched code branch is covered by
> > regression tests. I think the memory leak issue is independent of the
> > computational errors we've observed.
> > So, I'm going to push and
On Fri, Jan 14, 2022 at 7:30 PM Thomas Munro wrote:
> On Fri, Jan 14, 2022 at 4:35 PM Andres Freund wrote:
> > The more I think about it, the less I see why we *ever* need to re-arm the
> > latch in pq_check_connection() in this approach. pq_check_connection() is
> > only
> > used from from
Hi,
We see a successful compile and then a failure to read the file while
linking. We see that the animal got into that state recently and then
fixed itself, and now it's back in that state. I don't know if it's
significant, but it happened to fix itself when a configure change
came along,
At Mon, 14 Feb 2022 14:40:22 +0900 (JST), Kyotaro Horiguchi
wrote in
> It doesn't apply even on pg13 (due to LSN_FORMAT_ARGS). I will make
> the per-version patches if you are fine with this.
Oops! I forgot to rename the patch to avoid confusion on CF-bots.
I'll resend new version soon to
On Thu, Feb 10, 2022 at 11:38:44AM +0900, Michael Paquier wrote:
> The proposed change is incomplete anyway once you consider this
> argument. First, there is no need to set up those fields in
> EndPrepare() anymore if there is no origin data, no? It would be good
> to comment that these are
For a very long time, the AIX 7.1 buildfarm members (sungazer and tern)
have complained about
ip.c:236:17: warning: conversion from 'long unsigned int' to 'uchar_t' {aka
'unsigned char'} changes value from '1025' to '1' [-Woverflow]
I'd ignored this as not being very interesting, but I got
On Mon, Feb 14, 2022 at 6:10 AM Tom Lane wrote:
> Another thing I've been ignoring on the grounds that I-don't-feel-
> like-fixing-this is that the Solaris and OpenIndiana machines whine
> about every single reference from loadable modules to the core code,
> eg this from haddock while building
At Fri, 11 Feb 2022 01:00:03 +0900, Fujii Masao
wrote in
>
>
> On 2022/02/09 11:52, Kyotaro Horiguchi wrote:
> > 0001: The fix of CreateRestartPoint
>
> This patch might be ok for the master branch. But since concurrent
> checkpoint and restartpoint can happen in v14 or before, we would need
On Sun, Feb 13, 2022 at 5:49 PM John Naylor
wrote:
>
> On Sun, Feb 13, 2022 at 3:13 PM Bharath Rupireddy
> wrote:
> >
> > Hi,
> >
> > In the code comments, it is being used as "Startup process" in the
> > middle of the sentences whereas in most of the places "startup
> > process" is used. Also,
On Sun, Feb 13, 2022 at 9:56 PM Robert Haas wrote:
>
> On Sun, Feb 13, 2022 at 1:34 AM Dilip Kumar wrot>
> > test4:
> > 32 GB shared buffers, template DB size = 10GB, dirty shared buffer=70%
> > Head: 47656 ms
> > Patch: 79767 ms
>
> This seems like the most surprising result of the bunch. Here,
On Sat, Feb 12, 2022 at 12:10:46PM -0800, Andres Freund wrote:
> Tap tests currently are always executed in the source directory above t/, as
> far as I can tell. So I don't think VPATH/non-VPATH or windows / non-windows
> would break using a relative reference from the test dir.
That would mean
At Mon, 14 Feb 2022 14:42:18 +0900 (JST), Kyotaro Horiguchi
wrote in
> I'll resend new version soon to avoid the confusion..
In this version , 0001 gets one fix and two comment updates.
-* Archive recovery have ended. Crash recovery ever after
should always
+*
78 matches
Mail list logo
