On Tue, Oct 8, 2019 at 11:09 PM Amit Kapila wrote:
> On Sat, Oct 5, 2019 at 3:10 AM Andres Freund wrote:
> > On 2019-10-04 17:08:29 -0400, Tom Lane wrote:
> > > Andres Freund writes:
> > > > On 2019-10-04 16:31:29 -0400, Bruce Momjian wrote:
> > > >> Yeah, it is certainly weird that you have to
On Thu, Oct 17, 2019 at 3:52 PM Thomas Munro wrote:
> I rebased this patch, and tweaked get_collation_action_version() very
> slightly so that you get collation version change detection (of the
> ersatz kind provided by commit d5ac14f9) for the default collation
> even when not using
On Fri, Oct 18, 2019 at 1:21 AM Tom Lane wrote:
> Thomas Munro writes:
> > On Tue, Oct 15, 2019 at 1:55 PM Tom Lane wrote:
> >> and now prairiedog has shown it too:
> >> https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=prairiedog&dt=2019-10-14%2021%3A4
On Fri, Jun 21, 2019 at 6:52 AM Andres Freund wrote:
> On 2019-06-20 14:20:27 -0400, Robert Haas wrote:
> > On Tue, Jun 18, 2019 at 9:08 PM Thomas Munro wrote:
> > > Perhaps also the number of slots per backend should be dynamic, so
> > > that you have the option to i
Hi,
While working on some slides explaining EXPLAIN, I couldn't resist the
urge to add the missing $SUBJECT. The attached 0001 patch gives the
following:
Gather ... time=0.146..33.077 rows=1 loops=1)
Workers Planned: 2
Workers Launched: 2
Buffers: shared hit=4425
-> Parallel Seq Scan o
ojects propose to add new smgr
implementations alongside md.c in order to use bufmgr.c for more kinds
of files, but it seems entirely bogus to extend the unused smgr type
to cover those.
--
Thomas Munro
https://enterprisedb.com
0001-Remove-the-vestigial-smgr-type.patch
Description: Binary data
On Thu, Feb 28, 2019 at 7:08 PM Tom Lane wrote:
> Thomas Munro writes:
> > Motivation: A couple of projects propose to add new smgr
> > implementations alongside md.c in order to use bufmgr.c for more kinds
> > of files, but it seems entirely bogus to extend the unused s
On Thu, Feb 28, 2019 at 7:37 PM Tom Lane wrote:
> Thomas Munro writes:
> > On Thu, Feb 28, 2019 at 7:08 PM Tom Lane wrote:
> >> I agree that smgrtype as it stands is pretty pointless, but what
> >> will we be using instead to get to those other implementations?
&g
ithout too
much drama; that was one of the things that made me feel that the
magic database OID approach was acceptable at least in the short term.
--
Thomas Munro
https://enterprisedb.com
On Fri, Mar 1, 2019 at 4:09 AM Tom Lane wrote:
> Thomas Munro writes:
> > On Thu, Feb 28, 2019 at 7:37 PM Tom Lane wrote:
> >> Thomas Munro writes:
> >>> Our current thinking is that smgropen() should know how to map a small
> >>> number
On Fri, Mar 1, 2019 at 10:41 AM Shawn Debnath wrote:
> On Fri, Mar 01, 2019 at 10:33:06AM +1300, Thomas Munro wrote:
> > It doesn't make any sense to put things like clog or any other SLRU in
> > a non-default tablespace though. It's perfectly OK if not all smgr
> &
On Fri, Mar 1, 2019 at 11:31 AM Shawn Debnath wrote:
> On Thu, Feb 28, 2019 at 02:08:49PM -0800, Andres Freund wrote:
> > On 2019-03-01 09:48:33 +1300, Thomas Munro wrote:
> > > On Fri, Mar 1, 2019 at 7:24 AM Andres Freund wrote:
> > > > On 2019-02-28
athnodes.
>>
>> I had a look on this.
>>
>> The name "index skip scan" is a different feature from the
>> feature with the name on other prodcuts, which means "index scan
>> with postfix key (of mainly of multi column key) that scans
>> ignorin
On Thu, Feb 28, 2019 at 10:43 AM Thomas Munro wrote:
> On Thu, Feb 28, 2019 at 10:27 AM Shawn Debnath wrote:
> > We had a quick offline discussion to get on the same page and we agreed
> > to move forward with Andres' approach above. Attached is patch v10.
> > Here&#
th 128 MB shared_buffers, we will use an extra 16kB (one extra
> byte to represent the owner). I am hesitant to change this right now
> unless folks feel strongly about it.
>
> If so, I would combine the type and owner by splitting it up in 4 bit
> chunks, allowing for 16 request types and 16 smgrs. This change would
> only apply for the in-memory queue. The code and functions would
> continue using the enums.
+1
--
Thomas Munro
https://enterprisedb.com
n the
forseeable future). Now perhaps the single enum idea was going to
involve explicit values that encode the two values SYNC_REQ_CANCEL_MD
= 0x1 | (0x04 << 4) so you could still extract the requester part, but
that's just the same thing with uglier code.
--
Thomas Munro
https://enterprisedb.com
te your
own pseudo-filesystem that stores a whole PostgreSQL cluster inside
big data files or raw partitions. Luckily we don't need to tackle
such mountainous terrain to avoid the page cache, today.
--
Thomas Munro
https://enterprisedb.com
rallelism (TPC-H etc) but don't do any writes, or good for testing
writes (TPC-B etc) but don't do any parallelism. I'm going to
experiment with the "SIBENCH" approach from the Cahill paper and see
where that leads.
--
Thomas Munro
https://enterprisedb.com
Hash pathnames and
map to an array of lwlocks + error flags. Any process trying to sync
a file must hold the lock and check for a pre-existing error flag.
Now a checkpoint cannot succeed if any backend has recently decided to
panic. You could skip that if data_sync_retry = on.
--
Thomas Munro
https://enterprisedb.com
On Wed, Jan 23, 2019 at 11:23 AM Thomas Munro
wrote:
> On Wed, Jan 23, 2019 at 4:07 AM Tom Lane wrote:
> > The whole thing reminds me of the recent bug #15598:
> >
> > https://www.postgresql.org/message-id/87k1iy44fd.fsf%40news-spur.riddles.org.uk
>
> Yeah, if errors g
On Tue, Mar 5, 2019 at 10:08 AM Tom Lane wrote:
> I wrote:
> > Thomas Munro writes:
> >> That suggests that we could perhaps handle ECONNRESET both at startup
> >> packet send time (for certificate rejection, eelpout's case) and at
> >> initial query send
On Tue, Mar 5, 2019 at 11:35 AM Tom Lane wrote:
> Thomas Munro writes:
> > OK, here's something. I can reproduce it quite easily on this
> > machine, and I can fix it like this:
>
> >libpq_gettext("could not send startup packet: %s\n"),
> >
reeBSD system (note that fsync is disabled so
it's not that -- it must be bogus queue-related CPU?)
--
Thomas Munro
https://enterprisedb.com
ng on master and REL_11_STABLE:
https://buildfarm.postgresql.org/cgi-bin/show_failures.pl?max_days=200&branch=&stage=sslCheck&filter=Submit
Disappointingly, that turned out to be just because 10 and earlier
didn't care what the error message said.
--
Thomas Munro
https://enterprisedb.com
On Wed, Mar 6, 2019 at 3:33 AM Tom Lane wrote:
> Thomas Munro writes:
> > Disappointingly, that turned out to be just because 10 and earlier
> > didn't care what the error message said.
>
> That is, you can reproduce the failure on old branches? That lets
> out a h
are added we don't leave them
uninitialised. I like the syntax "FileTagData tag = {{0}}".
(Unfortunately extra nesting required here because first member is a
struct, and C99 doesn't allow us to use empty {} like C++, even though
some versions of GCC accept it. Rats.)
--
Thomas Munro
https://enterprisedb.com
On Wed, Mar 6, 2019 at 6:07 AM Tom Lane wrote:
> Thomas Munro writes:
> > You can see that poll() already knew the other end had closed the
> > socket. Since this is clearly timing... let's see, yeah, I can make
> > it fail every time by adding sleep(1) before the com
On Wed, Mar 6, 2019 at 7:05 AM Thomas Munro wrote:
> On Wed, Mar 6, 2019 at 6:07 AM Tom Lane wrote:
> > Annoying. I'd be happier about writing code to fix this if I could
> > reproduce it :-(
>
> Hmm. Note that eelpout only started doing it with OpenSSL 1.1.1.
Ble
.7 ./test.py --client
Sending A...
2
Sending B...
2
Sending C...
2
This is the server saying goodbye
So... can anyone tell us what happens on Windows?
(A secondary question might be what happens if the server and client
are on different machines since I guess it could be different?)
--
Thomas
e missing documentation.
Hmm, yeah. I wish these were alphabetised, I wish there was an
automated warning about this, I wish these tranches were declared a
better way that by adding code in RegisterLWLockTranches().
--
Thomas Munro
https://enterprisedb.com
missing-tranches.patch
Description: Binary data
On Wed, Mar 6, 2019 at 4:07 PM Shawn Debnath wrote:
> On Wed, Mar 06, 2019 at 11:13:31AM +1300, Thomas Munro wrote:
> > So... can anyone tell us what happens on Windows?
> C:\Users\Shawn Debnath\Desktop>c:\Python27\python.exe tmunro-ssl-test.py
> --client
> Sending
On Thu, Feb 28, 2019 at 7:02 PM Thomas Munro wrote:
> The type smgr has only one value 'magnetic disk'. ~15 years ago it
> also had a value 'main memory', and in Berkeley POSTGRES 4.2 there was
> a third value 'sony jukebox'. Back then, all tables had an
On Wed, Mar 6, 2019 at 6:16 AM Shawn Debnath wrote:
> On Tue, Mar 05, 2019 at 11:53:16AM -0500, Tom Lane wrote:
> > Thomas Munro writes:
> > > Why do we need to include fmgr.h in md.h?
> >
> > More generally, any massive increase in an include file's inclusi
it's for, rather than when to set it?
To make clear that it's a performance setting, not a safety one.
"Setting this option to off may increase
performance on copy-on-write filesystems."
--
Thomas Munro
https://enterprisedb.com
On Fri, Mar 8, 2019 at 12:35 PM Jerry Jelinek wrote:
> On Thu, Mar 7, 2019 at 3:09 PM Thomas Munro wrote:
>> My understanding is that it's not really the COW-ness that makes it
>> not necessary, it's the fact that fdatasync() doesn't do anything
>> different
/* without "force" flag raise exception immediately, or after
5 minutes */
Normally we call it an "error", not an "exception".
--
Thomas Munro
https://enterprisedb.com
id NOTIFY/LISTEN/UNLISTEN commands
NOTIFY notify_async2;
NOTIFY notify_async2, '';
--
Thomas Munro
https://enterprisedb.com
Try this first:
cd src/test/modules/test_shm_mq
make install
--
Thomas Munro
https://enterprisedb.com
ith exit code 1
Hmm, I don't know actually know why test_shm_mq_main() ends with
proc_exit(1) instead of 0. It's possible that it was written when the
meaning of bgworker exit codes was still being figured out, but I'm
not sure...
>> Works, thank you Thomas! I have spent m
ECKPOINT
Time: 101.848 ms
Patched:
postgres=# checkpoint;
CHECKPOINT
Time: 1.851 ms
--
Thomas Munro
https://enterprisedb.com
0001-Use-condition-variables-to-wait-for-checkpoints.patch
Description: Binary data
x27;re restarting the timeout clock every time through the loop
without adjustment, and I think that's not a good choice: spurious
wake-ups cause bonus waiting.
--
Thomas Munro
https://enterprisedb.com
s patch is to force the checkpointer to open and close
the files every time which seems OK to me. I know Andres wants to
make a pass through it too.
[1]
https://www.postgresql.org/message-id/CA%2BhUKGKN42jB%2BubCKru716HPtMbahdia39GwG5pLgWLMZ_y1ng%40mail.gmail.com
--
Thomas Munro
https://enterprisedb.com
ted by ckpt_lck, and
pushed.
There are some other things like this in the tree (grepping for
poll/pg_usleep loops finds examples in xlog.c, standby.c, ...). That
might be worth looking into.
--
Thomas Munro
https://enterprisedb.com
On Mon, Mar 4, 2019 at 10:17 AM Thomas Munro wrote:
> On Thu, Oct 11, 2018 at 10:15 AM Kevin Grittner wrote:
> > It applies and builds clean, it passed make world with cassert and TAP
> > tests, and I can't see any remaining flaws. This is true both of just
> > the 00
On Wed, Mar 6, 2019 at 9:21 AM Tom Lane wrote:
> Thomas Munro writes:
> > Bleugh. I think this OpenSSL package might just be buggy on ARM. On
> > x86, apparently the same version of OpenSSL and all other details of
> > the test the same, I can see that SSL_connect() retu
On Sun, Mar 17, 2019 at 2:00 AM Thomas Munro wrote:
> On Wed, Mar 6, 2019 at 9:21 AM Tom Lane wrote:
> > Yeah, I've still been unable to reproduce even with the sleep idea,
> > so eelpout is definitely looking like a special snowflake from here.
> > In any case, the
On Sun, Mar 17, 2019 at 5:53 PM Arseny Sher wrote:
> Thomas Munro writes:
>
> > 1. Why does pmdie()'s SIGTERM case terminate parallel workers
> > immediately? That breaks aborts running parallel queries, so they
> > don't get to end their work normally.
> &
the question is: is there a way for send() to report an error,
but recv() to report EAGAIN so we finish up blocked? I can't
immediately see how that could happen.
> My current feeling is that this is OK to put in HEAD but I think the
> risk-reward ratio isn't very good for the back branches. Even with
> an OpenSSL version where this makes a difference, the problematic
> behavior is pretty hard to hit. So I'm a bit inclined to do nothing
> in the back branches.
Shouldn't we also back-patch the one-line change adding
pqHandleSendFailure()? Then eelpout would be green for REL_11_STABLE,
and any future buildfarm animals (probably underpowered ones with
wonky scheduling) when they are eventually upgraded to OpenSSL 1.1.1.
--
Thomas Munro
https://enterprisedb.com
On Tue, Mar 19, 2019 at 12:25 PM Thomas Munro wrote:
> 2. Linux, FreeBSD and Darwin gave slightly different error sequences
> when writing after the remote connection was closed (though I suspect
> they'd behave much the same way for a connection to a remote host),
> but all all
On Tue, Mar 19, 2019 at 12:44 PM Tom Lane wrote:
> Thomas Munro writes:
> > Shouldn't we also back-patch the one-line change adding
> > pqHandleSendFailure()?
>
> As I said before, I don't like that patch: at best it's an abuse of
> pqHandleSendFailure, be
On Sat, Feb 16, 2019 at 10:57 PM Thomas Munro
wrote:
> Yeah. This coding is ugly and StringInfo would be much nicer.
> Thinking about that made me realise that the proposed SRV case should
> also handle multiple SRV records by building a multi-URL string too
> (instead of just taki
ta is over.
This plan looks good to me.
--
Thomas Munro
https://enterprisedb.com
on HOT updates.
Still, those are just minor nitpicks, and I don't expect that to affect the
quality of the patch implementation.
Good show, gents!
--
Shaun M Thomas - 2ndQuadrant
PostgreSQL Training, Services and Support
shaun.tho...@2ndquadrant.com | www.2ndQuadrant.com
a" test is ssl, but that was already working.
--
Thomas Munro
https://enterprisedb.com
ldap-on-macports.patch
Description: Binary data
On Tue, Mar 19, 2019 at 9:01 PM Thomas Munro wrote:
> I'd like to commit this soon.
Done, after some more comment adjustments. Thanks Daniel and Graham
for your feedback!
--
Thomas Munro
https://enterprisedb.com
On Mon, Feb 4, 2019 at 8:41 PM Andres Freund wrote:
> On 2018-09-19 13:58:36 +1200, Thomas Munro wrote:
>
> > +/*
> > + * Advance nextFullXid to the value after a given xid. The epoch is
> > inferred.
> > + * If lock_free_check is true, then the caller must
On Mon, Mar 25, 2019 at 5:01 PM Thomas Munro wrote:
> New version attached. I'd like to commit this for PG12.
Here is a follow-up sketch patch that shows FullTransactionId being
used in the transaction stack, so you can call eg
GetCurrentFullTransactionId(). A table access method c
On Thu, Mar 21, 2019 at 4:39 PM Tom Lane wrote:
> Thomas Munro writes:
> > Peter E added some nice tests for LDAP and Kerberos, but they assume
> > you have Homebrew when testing on a Mac. Here's a patch to make them
> > work with MacPorts too (a competin
a Deathstation 9000 might allocate 8 billion more xids
with all the necessary auto-vacuuming to allow that before scheduling
you back onto the CPU. Admittedly, I haven't though about this very
deeply at all, there may be a simple and correct way to do it.
--
Thomas Munro
https://enterprisedb.com
On Wed, Mar 13, 2019 at 2:27 PM Thomas Munro wrote:
> [...] Aside from some refactoring which I
> think looks good anyway and prepares for future patches, the main
> effect of this patch is to force the checkpointer to open and close
> the files every time which seems OK to me.
I&#x
On Tue, Mar 26, 2019 at 12:58 PM Thomas Munro wrote:
> ... I think you could probably reclaim that space by
> using a more compact representation of vacuumFlags, overflowed,
> delayChkpt, nxids (it's funny, the comment says "as tightly as
> possible", which clearly
ch == now_epoch && xid > now_epoch_last_xid))
+ || (xid_epoch == now_epoch && xid >= now_epoch_last_xid))
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("transaction
On Wed, Mar 27, 2019 at 7:55 PM Thomas Munro wrote:
> If you keep asking for txid_status(txid_current() + 1) in new
> transactions, you eventually hit:
>
> ERROR: could not access status of transaction 32768
> DETAIL: Could not read from file "pg_xact/" at
On Tue, Mar 26, 2019 at 12:58 PM Thomas Munro wrote:
> On Tue, Mar 26, 2019 at 3:23 AM Heikki Linnakangas wrote:
> > Looks good.
I did some testing and proof-reading and made a few minor changes:
* I tidied up the code that serialises transaction state. It was
already hammering r
concrete
reason not to abuse
the parameter. :shipit:
Vaya con Queso
--
Shaun M Thomas - 2ndQuadrant
PostgreSQL Training, Services and Support
shaun.tho...@2ndquadrant.com | www.2ndQuadrant.com
o weird
new problem in heap_compute_xid_horizon_for_tuples() which I assumed
to be somehow my fault due to the mention of xid horizons, but I
eventually realised that master was broken on that machine and
followed that up elsewhere. Phew.
Thanks for the reviews! Pushed.
--
Thomas Munro
https://enterprisedb.com
On Fri, Mar 29, 2019 at 8:59 AM Robert Haas wrote:
> On Tue, Mar 26, 2019 at 3:24 PM Jerry Jelinek
> wrote:
> > The latest patch is rebased, builds clean, and passes some basic testing.
> > Please let me know if there is anything else I could do on this.
>
> I agree wit
On Fri, Mar 29, 2019 at 10:47 AM Thomas Munro wrote:
> On Fri, Mar 29, 2019 at 8:59 AM Robert Haas wrote:
> > On Tue, Mar 26, 2019 at 3:24 PM Jerry Jelinek
> > wrote:
> > > The latest patch is rebased, builds clean, and passes some basic testing.
> > > Please
On Wed, Dec 13, 2017 at 5:01 PM Thomas Munro
wrote:
> On Wed, Dec 13, 2017 at 4:24 PM, Andres Freund wrote:
> > Hi,
> >
> > Buildfarm animal thrips just failed with a curious error:
> > https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=thrips&a
know so someone else can pick
it up.
--
Shaun M Thomas - 2ndQuadrant
PostgreSQL Training, Services and Support
shaun.tho...@2ndquadrant.com | www.2ndQuadrant.com
Hello,
Patch 0001 gets rid of the unconditional lseek() calls for SLRU I/O,
as a small follow-up to commit c24dcd0c. Patch 0002 gets rid of a few
places that usually do a good job of avoiding lseek() calls while
reading and writing WAL, but it seems better to have no code at all.
--
Thomas
no IO
concurrency at all (a single old school spindle), just because the OS
will chew on that queue in the background while we're also doing
stuff, which is probably what that "+ 10" is expressing. But that
seems to apply to bitmap heapscan too, doesn't it?
--
Thomas Munro
https://enterprisedb.com
override
errors automatically as follows, though that is proving trickier to
get working. Example:
postgres=# SUDO DROP TABLE PG_DATABASS;
NO CARRIER
--
Thomas Munro
https://enterprisedb.com
0001-Add-do-what-I-mean-mode-to-psql.patch
Description: Binary data
LueN9rofAA3HHSYikW-Zw%40mail.gmail.com
I see that its first failure was after commit 558a9165e0 (along with others).
--
Thomas Munro
https://enterprisedb.com
anups, documentation, and automatic back traces from assertion failures.
Now works out of the box on FreeBSD. The assertion thing is a nice touch.
I wonder if it'd make sense to have a log_min_backtrace GUC that you
could set to error/fatal/panicwhatever (perhaps in a later patch).
--
Thomas Munro
https://enterprisedb.com
On Fri, Jul 19, 2019 at 3:00 PM Amit Kapila wrote:
> On Thu, Jul 18, 2019 at 7:15 PM Tom Lane wrote:
> > Thomas Munro writes:
> > > Hmm, so something like a new argument "bool final" added to the
> > > ExecXXXShutdown() functions, which receives false in
On Tue, Jul 16, 2019 at 1:11 AM Robert Haas wrote:
> On Fri, Jul 12, 2019 at 11:03 PM Thomas Munro wrote:
> > I pushed this too. It's a separate commit, because I think there is
> > at least a theoretical argument that it should be back-patched. I'm
> > no
ell_201907_unique_idx ON
> child.unused0_huawei_umts_nodeb_locell_201907 USING btree ...
> ... I've set
> max_parallel_workers_per_gather=0, ...
Just by the way, parallelism in CREATE INDEX is controlled by
max_parallel_maintenance_workers, not max_parallel_workers_per_gather.
--
Thomas Munro
https://enterprisedb.com
aster dies while they are waiting for workers; most other places
(probably every other backend in your cluster) just quietly exit.
That tells us something about what's happening, but on its own doesn't
tell us that parallelism plays an important role in the failure mode.
--
Thomas Munro
https://enterprisedb.com
On Wed, Jul 24, 2019 at 10:03 AM Thomas Munro wrote:
> > edata =
> If you have that core, it might be interesting to go to frame 2 and
> print *edata or edata->saved_errno. ...
Rats. We already saw that it's optimised out so unless we can find
that somewhere else
On Wed, Jul 24, 2019 at 10:42 AM Justin Pryzby wrote:
> On Wed, Jul 24, 2019 at 10:03:25AM +1200, Thomas Munro wrote:
> > On Wed, Jul 24, 2019 at 5:42 AM Justin Pryzby wrote:
> > > #2 0x0085ddff in errfinish (dummy=) at
> > > elog.c:555
> > >
y a panic if a
checkpoint happens to run exactly at a moment where the create index
hasn't yet written the byte that breaks the camel's back, but the
checkpoint pushes it over edge in one of these places where it panics
on failure.
--
Thomas Munro
https://enterprisedb.com
On Wed, Jul 24, 2019 at 10:11 AM Tom Lane wrote:
> Thomas Munro writes:
> > *I suspect that the only thing implicating parallelism in this failure
> > is that parallel leaders happen to print out that message if the
> > postmaster dies while they are waiting for worke
animals on one
kernel, and need to increase net.core.rmem_max and
net.core.rmem_default (or maybe the write side variants, or both, or
something like that).
--
Thomas Munro
https://enterprisedb.com
should do that too. I don't know how to write Perl but I'll try...
--
Thomas Munro
https://enterprisedb.com
On Wed, Jul 24, 2019 at 3:52 PM Thomas Munro wrote:
> I guess we should do that too. I don't know how to write Perl but I'll try...
Does this look about right?
--
Thomas Munro
https://enterprisedb.com
wait-for-slapd.patch
Description: Binary data
, and the
> sleep phases make use of Time::HiRes::usleep.
Thanks, here's v2.
--
Thomas Munro
https://enterprisedb.com
wait-for-slapd-v2.patch
Description: Binary data
version mentioned in tools/pgindent/README).
Fixed.
--
Thomas Munro
https://enterprisedb.com
wait-for-slapd-v3.patch
Description: Binary data
is unsigned. If we had a SQL type for 64 bit
xids, it should be convertible to xid, and the reverse conversion
should require a more complicated dance. Of course we can't casually
change txid_current() without annoying people who are using it, so
perhaps if we invent a new SQL
the past
> >> and it looked like a mess.
>
> > I assume Thomas was thinking more of another bespoke type like xid, just
> > wider. There's some notational advantage in not being able to
> > immediately do math etc on xids.
>
> Well, we could invent an x
postgresql-9.6 and postgresql-9.6-dbg installed you could run "gdb
/usr/lib/postgresql/9.6/bin/postgres" and then "disassemble
ExecHashJoin". The code at "<+1442>" (0x5a2) is presumably calling
free or some other libc thing (though I'm surprised not to see an
intervening palloc thing).
--
Thomas Munro
https://enterprisedb.com
until a checkpoint says they existed at that point in the WAL stream.
Does this make sense?
BTW we could probably use posix_fallocate() instead of writing zeroes;
I think Andres mentioned that recently. I see also that someone tried
that for WAL and it got reverted back in 2013 (commit
b1892aaeaaf34d8d1637221fc1cbda82ac3fcd71, I didn't try to hunt down
the discussion).
[1] https://lkml.org/lkml/2012/9/3/83
[2]
https://github.com/torvalds/linux/commit/b71fc079b5d8f42b2a52743c8d2f1d35d655b1c5
--
Thomas Munro
https://enterprisedb.com
!
Pushed, thanks.
--
Thomas Munro
https://enterprisedb.com
laint[2]), but I ran
out of time for debugging adventures today.
[1]
https://www.postgresql.org/message-id/CA%2BhUKGJyqDp9FZSHLTjiNMcz-c6%3DRdStB%2BUjVZsR8wfHnJXy8Q%40mail.gmail.com
[2]
https://www.postgresql.org/message-id/flat/86137f17-1dfb-42f9-7421-82fd786b04a1%40anayrat.info
--
Thomas Munro
https://enterprisedb.com
reconsider.
FYI This is now on LLVM's release_90 branch, due out on August 28.
--
Thomas Munro
https://enterprisedb.com
ot;,
> less than 400 hundred instances, and backpatch to all supported
> versions:-(
I would just #undef Min for our small number of .cpp files that
include LLVM headers. It's not as though you need it in C++, which
has std::min() from .
--
Thomas Munro
https://enterprisedb.com
On Sat, Jul 27, 2019 at 7:12 PM Thomas Munro wrote:
> On Sat, Jul 27, 2019 at 7:06 PM Fabien COELHO wrote:
> > Maybe we should consider doing an explicit bug report, but I would not bet
> > that they are going to fold… or fixing the issue pg side, eg "pg_Min",
> >
ake
check. You can see it on Linux with ./configure ...
CFLAGS="-DWAIT_USE_POLL".
--
Thomas Munro
https://enterprisedb.com
ing ready to tell us if something else breaks. Personally, I don't
see any reason why should entertain a request
to change their variable names to avoid our short common word macros
that aren't even all-caps, but if someone asks them and they agree to
do that before the final 9.0 rel
On Mon, Jul 29, 2019 at 9:55 AM Tom Lane wrote:
> Thomas Munro writes:
> > Let's just commit the #undef so that seawasp is green and back to
> > being ready to tell us if something else breaks.
>
> +1. I was afraid that working around this would be impossibly
> pai
701 - 800 of 4331 matches
Mail list logo
