Обсуждение: Re: phone

Here is Peter's analysis of our current 7.3 backlog:
http://candle.pha.pa.us/cgi-bin/pgpatcheshttp://candle.pha.pa.us/cgi-bin/pgpatches2

---------------------------------------------------------------------------


Well, since I'm complaining I might as well go through the patches files:

patches (48 hour queue)
=======

* Whatever ... JDBC build patch

Has been rejected 5 times before and ain't gonna change now.

* Patch to add CREATE OPERATOR CLASS

I don't like that syntax.  The internal functionality is probably OK.  I
suggest the author resubmit the last patch and we can check it out.

* BUG and FIX: getTables - SYSTEM TABLE

Is this really a patch or just an observation?  The JDBC people should
know.

* ECPG patches: get descriptor NULL alloc, external names

Don't know.  (Although Cristof's ECPG patches tend to be good.)

* stupid patch of pg_dumplo

See Karel's comments.  Also, I'd like to consider pg_dumplo obsolete.

* Support for QNX6, POSIX IPC and PTHREAD-style locking

Patch needs serious cleanups in build system.  Should be split into two or
more patches for generic QNX6 support and for the locking and IPC changes.

* Numeric decimal digits and radix to DatabaseMetaData.getColumns()

Not sure how to treat "little patches" that don't explain what they do.

* Libpq support for precision and scale

Tom had lots of issues with it.  I don't like the patch at all.

* Patch to add Heimdal kerberos support, with patch

I'm working on that right now.

* Patch : Re: JDBC improvements

Check with JDBC developers.

* Heimdal Kerberos 5 support in 7.1.3

Working on that.

* [HACKERS]

More QNX6 stuff that's even less polished.  These two patches will need to
be merged I guess.

* PQunescapeBytea libpq function

Seems reasonable.  I think there's an PQescapeBytea function already.  So
it would make sense.  Haven't checked the actual behaviour though.

* psql \timing patch

Looks OK.  Will check out closer soon.

* Updated TODO item

Have the author throw the latest patch up for review.

* Undocumented feature costs a lot of performance in COPY

Not sure what this is supposed to do.

* Makefile.aix patch for shlib's

Should probably have been applied already?

* Locale support for postgresql regex (src)

Does not support multibyte.  But the issue might be worth a TODO item.
("Make regex character classes respect locale")

* COPY when 'filename' is a directory

Seems OK

* Problem in ResultSet#getTimestamp() of 7.2b4

Don't know.

* postgresql-7.2b3-betterquote.patch

This affects the Python interface.  Ask D'Arcy.

* contrib: int_array_aggregator() int_array_enum()

I can't understand the description given.

* Locale support for postgresql regex (src)

Does not support multibyte.  But the issue might be worth a TODO item.
("Make regex character classes respect locale")

* enable debugging in jdbc

OK if JDBC people agree.

* Fix for non-blocking connections in libpq

Run this by the list one more time for review.

* GiST code cleanup

Run it by someone who knows.

* contrib/DBMirror (Replication)

I'd be hesitant to add more maybe-replication solutions to contrib.

* JDBC Connection startup cleaned up

Let JDBC folks handle this.

* Cygwin plperl patch

OK to check in.

* hashing improvements

Tom Lane didn't protest the last patch, so it seems OK.

* pg_ctl - tighten command parameter checking

Not portable.  But you might want to add to TODO something like, In shell
scripts make sure that options requiring arguments get arguments, to
prevent overshifting.

* Make equals sign optional in CREATE DATABASE WITH param = 'val'

Seems OK.

* Updates errors_zh_TW.properties of JDBC driver

OK

* TOAST-slicing: substring access methods for TOASTed values.

I'd like to see the description posted to the list so everyone knows what
we're getting.  Maybe this was already done, but I can only find little
feedback.

* Added Encoding

Tatsuo liked it.  That's good enough for me.  But he wanted to see some
documentation updates.


patches2 (unfiltered queue)
========

* nocreatetable for 7.1.2

Obsoleted by schema work (fingers crossed)

* KSQO parameter

Evidently not wanted by Tom Lane

* Index location patch for review

Objected to by Tom Lane on -hackers today.

* CLUSTER TODO item

Ask Tom.  Probably should look at the whole patch at least once more.

* Combining chars in psql

Well, it says "pre-patch".

* Select * from cursor foo

No idea.

* IPv6 Support for INET/CIDR types

Currently under discussion on -hackers.  Needs work on output format.

* Efficient slicing/substring of TOAST values

See under "patches".

* Proposed new create command, CREATE OPERATOR CLASS

See under "patches".

* Storage Location Patch Proposal for V7.3

This is just a proposal, nowhere near final.

* DBD::Pg BYTEA Character Escaping

This does not contain a patch and we don't maintain DBD::Pg anyway.

* Added Encoding

See under "patches".

* Locale timings

Well, this is my material, which I will get to in a few weeks perhaps.

* Libpq support for precision and scale

See above under "patches".

* Backend error message style issues

This is my material again.

* date formatting and tab-complete patch

Let's hang on to that one for a little.  It's a good idea, but it needs
some reformatting.

* Pl/Tcl problem

Andreas Zeugwetter had a patch for this, which we should work with.

* Problem compiling postgres sql --with-tcl

same

* system catalog relation of a table and a serial sequence

I think we found that dumping dependencies by default is not such a hot
idea after all.  The original patch had problems, and I didn't see a good
one after that.  There were some issues raised that we might want to
discuss again.

* Thoughts on the location of configuration files, how about this

We still have the problem of securing the moved config files.

* Submittal for JDBC Pooling driver (For 7.3)

Better ask the JDBC guys.

* patch: INSERT INTO t VALUES (a, b, ..., DEFAULT, ...)
* TODO question

These appear to go together.  Not sure what the last patch really is.  Ask
the author.

* Casting Varchar to Numeric

... is just a bad idea.

* Updated TODO item

See above.

* Undocumented feature costs a lot of performance in COPY

Not sure what this is supposed to do.

* Some interesting results from tweaking spinlocks

This is only a suggestion, not a patch.

* --with-tcl build on AIX (and others) fails

See Tcl above.  I can take care of those.

* Makefile.aix patch for shlib's

Should probably have been applied already?

* notify support for DBD::Pg

Does not live here.

* postgresql-7.2b3-betterquote.patch

This affects the Python interface.  Ask D'Arcy.

* Locale support for postgresql regex (src)

Does not support multibyte.  But the issue might be worth a TODO item.
("Make regex character classes respect locale")

* LISTEN/NOTIFY support in JDBC driver?

The message says it's already applied.

* guc

This patch was effectively rejected.  The TODO item "Add SET REAL_FORMAT
and SET DOUBLE_PRECISION_FORMAT using printf args" has effectively been
rejected with it.  What we want is to be able to print floating-point
numbers in a portable binary representation (sprintf("%a")) for the
purposes of pg_dump.

* contrib: int_array_aggregator() int_array_enum()

I can't understand the description given.

* Which perl5 Pg to use? Who maintains it?

We do.  At least.

* pl/pgsql Composite Parameter Question

Not a patch.  Maybe a TODO.

* enable debugging in jdbc

See under patches.

* libpq++

Not sure what this is doing in the patch queue.

* C++: PgDatabase or PgConnection and lifetime of the objects?

Same.

* Fix for non-blocking connections in libpq

Run this by the list one more time for review.

* libpq not reentrant

No patch there.  Maybe a TODO.

* ERROR (Bug?) in RULE processing ?

Don't know, but there's not a patch there.

* new module contrib/tree for 7.2 ?

Don't know the status.

* GiST code cleanup

Run it by someone who knows.

* contrib/DBMirror (Replication)

See under patches.

* JDBC Connection startup cleaned up

Let JDBC folks handle this.

* Cygwin plperl patch

OK to check in.


The remaining items are recent and possibly still under discussion.  I
don't know the latest status on them.

-  Bruce Momjian                        |  http://candle.pha.pa.us pgman@candle.pha.pa.us               |  (610)
853-3000+  If your life is a hard drive,     |  830 Blythe Avenue +  Christ can be your backup.        |  Drexel Hill,
Pennsylvania19026
 

Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Here is Peter's analysis of our current 7.3 backlog:

Some followup comments:

> * Patch to add CREATE OPERATOR CLASS
> I don't like that syntax.

Me either.  I'm quite certain that there was followup discussion to this
version of the patch; I don't recall if Bill had submitted a revised
patch, but this version should not be applied.

> * ECPG patches: get descriptor NULL alloc, external names

> Don't know.  (Although Cristof's ECPG patches tend to be good.)

Michael Meskes should have the responsibility to review and apply ecpg
patches (actually I think he did already apply everything from Cristof).

> * Support for QNX6, POSIX IPC and PTHREAD-style locking

> Patch needs serious cleanups in build system.  Should be split into two or
> more patches for generic QNX6 support and for the locking and IPC changes.

Agreed.  I would like to see more discussion about what sorts of
alternate locking/IPC mechanisms we want to support before, rather than
after, someone plugs in new code.  That should in turn lead to some kind
of implementation-independent API for the locking primitives, rather
than piling one set of implementation dependencies atop another, as
here.  This patch is a good starting point for that discussion --- but
I don't want to apply it as-is.

> * Undocumented feature costs a lot of performance in COPY

> Not sure what this is supposed to do.

The idea was to make the syntax of the COPY options a little more
consistent, which is fine --- but where's the documentation update
to go with this?  Ask submitter to resubmit with doco patch included.

> * contrib: int_array_aggregator() int_array_enum()

> I can't understand the description given.

Seems marginally useful as a contrib example ... though I agree the
README could be improved, and some attention to spelling would be good.

> * Locale support for postgresql regex (src)

> Does not support multibyte.  But the issue might be worth a TODO item.

Agreed, this cannot be applied as-is.

> * Fix for non-blocking connections in libpq

> Run this by the list one more time for review.

There was followup discussion on this when it was submitted; I don't
recall if we decided it was okay as-is.  I do know that I've been
unhappy with the nonblocking extension to libpq since day one.

> * hashing improvements

> Tom Lane didn't protest the last patch, so it seems OK.

This patch is OK to apply.

> * Updated TODO item [Gavin Sherry's patch to add OWNER option to CREATE DB]

> Have the author throw the latest patch up for review.

It looks okay to me, except for the // comments and a couple other
trivial details, and the lack of documentation patches; but basically
it's a good feature, and we need it so we can fix up pg_dumpall to not
assume database owners have createdb privilege.  I will take
responsibility for fixing and applying this one.

> * Make equals sign optional in CREATE DATABASE WITH param = 'val'

> Seems OK.

This is a subset of Gavin Sherry's patch above; we don't need it.


> patches2 (unfiltered queue)
> ========

> * CLUSTER TODO item

> Ask Tom.  Probably should look at the whole patch at least once more.

Gavin made clear that that patch wasn't ready to apply.  Remove from queue.

> * Select * from cursor foo

> No idea.

Not ready for prime time either.

> * Pl/Tcl problem
> Andreas Zeugwetter had a patch for this, which we should work with.
> * Problem compiling postgres sql --with-tcl
> same

Yes, I think Andreas' patch should be the starting point for these
issues.  Not sure if we were ready to apply it or not, though.

> * Casting Varchar to Numeric

> ... is just a bad idea.

At least till we have a distinction between explicitly and implicitly
invokable casts.  An explicit-only text->numeric cast would be fine with me.

> * guc

> This patch was effectively rejected.  The TODO item "Add SET REAL_FORMAT
> and SET DOUBLE_PRECISION_FORMAT using printf args" has effectively been
> rejected with it.  What we want is to be able to print floating-point
> numbers in a portable binary representation (sprintf("%a")) for the
> purposes of pg_dump.

Well, Peter may not like the concept but I still do.  I agree that this
particular patch was rejected.

> The remaining items are recent and possibly still under discussion.  I
> don't know the latest status on them.

I believe Fernando Nasser's "Allow arbitrary levels of
analyze/rewriting" is fine.  I wasn't happy with the details of Brent's
"Problem reloading regression database" patch; that still needs another
iteration.  Not sure about the rest.  Could you clean out the cruft
(non-patch items and stuff we've already agreed to reject) so it's
easier to see what still needs review?
        regards, tom lane

> > The remaining items are recent and possibly still under discussion.  I
> > don't know the latest status on them.
> 
> I believe Fernando Nasser's "Allow arbitrary levels of
> analyze/rewriting" is fine.  I wasn't happy with the details of Brent's
> "Problem reloading regression database" patch; that still needs another
> iteration.  Not sure about the rest.  Could you clean out the cruft
> (non-patch items and stuff we've already agreed to reject) so it's
> easier to see what still needs review?

Cleaned out:
http://candle.pha.pa.us/cgi-bin/pgpatches
http://candle.pha.pa.us/cgi-bin/pgpatches2

Patches list down to 33.  I know some of our special people will chime
in tomorrow with comments, e.g. jdbc, gist, python, etc.


--  Bruce Momjian                        |  http://candle.pha.pa.us pgman@candle.pha.pa.us               |  (610)
853-3000+  If your life is a hard drive,     |  830 Blythe Avenue +  Christ can be your backup.        |  Drexel Hill,
Pennsylvania19026
 

>From the jdbc side it looks like we are pretty much in sync here. I have
applied a few patches and am working on my queue which appears to be
more or less the same as yours.

BTW I got a message from Barry; he and his wife have a new baby and he
will be offline for a while.

Dave

-----Original Message-----
From: pgsql-hackers-owner@postgresql.org
[mailto:pgsql-hackers-owner@postgresql.org] On Behalf Of Bruce Momjian
Sent: Sunday, February 24, 2002 11:26 PM
To: Tom Lane
Cc: PostgreSQL-development
Subject: Re: [HACKERS] patch queue


> > The remaining items are recent and possibly still under discussion.

> > I don't know the latest status on them.
> 
> I believe Fernando Nasser's "Allow arbitrary levels of 
> analyze/rewriting" is fine.  I wasn't happy with the details of 
> Brent's "Problem reloading regression database" patch; that still 
> needs another iteration.  Not sure about the rest.  Could you clean 
> out the cruft (non-patch items and stuff we've already agreed to 
> reject) so it's easier to see what still needs review?

Cleaned out:
http://candle.pha.pa.us/cgi-bin/pgpatches
http://candle.pha.pa.us/cgi-bin/pgpatches2

Patches list down to 33.  I know some of our special people will chime
in tomorrow with comments, e.g. jdbc, gist, python, etc.


--  Bruce Momjian                        |  http://candle.pha.pa.us pgman@candle.pha.pa.us               |  (610)
853-3000+  If your life is a hard drive,     |  830 Blythe Avenue +  Christ can be your backup.        |  Drexel Hill,
Pennsylvania
19026

---------------------------(end of broadcast)---------------------------
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to majordomo@postgresql.org so that your
message can get through to the mailing list cleanly

Dave Cramer wrote:
> >From the jdbc side it looks like we are pretty much in sync here. I have
> applied a few patches and am working on my queue which appears to be
> more or less the same as yours.
> 
> BTW I got a message from Barry; he and his wife have a new baby and he
> will be offline for a while.

Yes, I heard.  I tried to reply but the email address was rejected.  Let
me know when you are done and I will delete all the jdbc patches, or I
can just delete all them now.  Let me know.

--  Bruce Momjian                        |  http://candle.pha.pa.us pgman@candle.pha.pa.us               |  (610)
853-3000+  If your life is a hard drive,     |  830 Blythe Avenue +  Christ can be your backup.        |  Drexel Hill,
Pennsylvania19026