Обсуждение: pg_dump fails

I'm running  postgresql 7.4.7 on x86 and have run into a problem trying
to export my database. When I use pg_dump --file=stats.out --format=t
printstats, (and I may have the syntax wrong), I get this error:
pg_dump: SQL command failed
pg_dump: Error message from server: ERROR:  did not find '}' at end of
input node
pg_dump: The command was: SELECT (SELECT usename FROM pg_user WHERE
usesysid = datdba) as dba, pg_encoding_to_char(encoding) as encoding,
datpath FROM pg_database WHERE datname = 'printstats'

I tried googling for an answer, and found a couple of references to
this error, but nothing that could help me. Can anyone help me out
here?
Thanks


Tech/Library Combo Lab Manager
Northwestern University
Office Tech MG49
mailto:lthurman@northwestern.edu
voice: 847-467-6565
pager: 847-536-0094

Lorenzo Thurman <lorenzo@thethurmans.com> writes:
> pg_dump: Error message from server: ERROR:  did not find '}' at end of
> input node
> pg_dump: The command was: SELECT (SELECT usename FROM pg_user WHERE
> usesysid = datdba) as dba, pg_encoding_to_char(encoding) as encoding,
> datpath FROM pg_database WHERE datname = 'printstats'

This suggests that there's something wrong with the pg_user view ---
specifically, that the stored view rule isn't in the format that the
backend expects.

I don't recall ever having seen that error in the field, only in
development builds that someone had been sloppy about fully rebuilding
after a CVS update (so that the code inside the server wasn't fully
self-consistent).  Where did you get your postgres executable from,
exactly?

Also, does "select * from pg_user" provoke the same error in every
database of your installation, or only this one?  If the latter,
it could be a data-corruption kind of problem.

            regards, tom lane

On Apr 18, 2005, at 4:46 p, Tom Lane wrote:


<excerpt>Also, does "select * from pg_user" provoke the same error in
every

database of your installation, or only this one?  If the latter,

it could be a data-corruption kind of problem.

</excerpt>

I only have one database right now. When I try "select * from
pg_user", I get this error:

ERROR:  did not find '}' at end of input node


I installed this from Gento's portage repository.


<smaller>"There are 10 types of people in this world: those who
understand binary, those who don't"

<italic>--Unknown</italic></smaller>


On Apr 18, 2005, at 4:46 p, Tom Lane wrote:

> Also, does "select * from pg_user" provoke the same error in every
> database of your installation, or only this one?  If the latter,
> it could be a data-corruption kind of problem.

I only have one database right now. When I try "select * from pg_user",
I get this error:
ERROR:  did not find '}' at end of input node

I installed this from Gento's portage repository.

"There are 10 types of people in this world: those who understand
binary, those who don't"
--Unknown

Lorenzo Thurman <lorenzo@thethurmans.com> writes:
>>> ERROR:  did not find '}' at end of input node

> I installed this from Gento's portage repository.

Gentoo has a history of supplying broken compilers, software, etc
... bleeding edge stuff tends to cut you occasionally :-(

You might try updating your compiler and then rebuilding PG.

            regards, tom lane

I'm trying that right now. I think there may be mis-match in the build
settings between upgrades of postgresql. The "USE" settings may be at
fault:


 - - pg-hier        : Enables recursive queries like Oracle's 'CONNECT
BY' feature.

 - - pg-vacuumdelay : Adds in the vacuum inter-page delay feature.

 - - pg-intdatetime : Enables --enable-integer-datetimes configure
option, which changes PG to use 64-bit integers for timestamp storage


I think these may have been changed since the original install.


On Apr 18, 2005, at 11:57 p, Tom Lane wrote:


<excerpt>Lorenzo Thurman <<lorenzo@thethurmans.com> writes:

<excerpt><excerpt><excerpt>ERROR:  did not find '}' at end of input
node

</excerpt></excerpt></excerpt>

<excerpt>I installed this from Gento's portage repository.

</excerpt>

Gentoo has a history of supplying broken compilers, software, etc

... bleeding edge stuff tends to cut you occasionally :-(


You might try updating your compiler and then rebuilding PG.


            regards, tom lane


</excerpt><smaller>I got an object

I was sure it was point

ClassCastException

-- <italic>Jens Alfke</italic></smaller>

I'm trying that right now. I think there may be mis-match in the build
settings between upgrades of postgresql. The "USE" settings may be at
fault:

  - - pg-hier        : Enables recursive queries like Oracle's 'CONNECT
BY' feature.
  - - pg-vacuumdelay : Adds in the vacuum inter-page delay feature.
  - - pg-intdatetime : Enables --enable-integer-datetimes configure
option, which changes PG to use 64-bit integers for timestamp storage

I think these may have been changed since the original install.

On Apr 18, 2005, at 11:57 p, Tom Lane wrote:

> Lorenzo Thurman <lorenzo@thethurmans.com> writes:
>>>> ERROR:  did not find '}' at end of input node
>
>> I installed this from Gento's portage repository.
>
> Gentoo has a history of supplying broken compilers, software, etc
> ... bleeding edge stuff tends to cut you occasionally :-(
>
> You might try updating your compiler and then rebuilding PG.
>
>             regards, tom lane
>
I got an object
I was sure it was point
ClassCastException
-- Jens Alfke

Lorenzo Thurman <lorenzo@thethurmans.com> writes:
> I'm trying that right now. I think there may be mis-match in the build
> settings between upgrades of postgresql. The "USE" settings may be at
> fault:

>   - - pg-hier        : Enables recursive queries like Oracle's 'CONNECT
> BY' feature.

[ rolls eyes... ]  Yup, that's Gentoo all right: throw in random patches
that have been rejected by the upstream developers.  Now that I think
about it, this failure is exactly what that patch is known to cause,
because it makes an incompatible change in Query structures and hence
in on-disk view rule representation.

> I think these may have been changed since the original install.

Go back to your prior setting, or even better stop using Gentoo's
hacked-up version.  I'm not sure why we even bother to answer support
requests from Gentoo users, when what they are using is not our
software but some randomly-modified variant.  I wonder what other
brokennesses Gentoo may be including ...

(Just for the record: I work for Red Hat, which has a rather different
notion of the level of reliability it wants to ship.  So take my opinion
with the appropriate grain of salt.  But I'm a mite ticked off at the
moment --- you're not the first person to have been bitten by this,
and you likely won't be the last, and I think it's entirely because
Gentoo has such a low quality standard for the patches they ship.)

            regards, tom lane

Thanks for the reply. I've tried recompiling with my install build
settings, but no luck. I've posted a message on the Gentoo forums.
Hopefully they will have an answer. If they do, I'll post back here for
future reference.

On Apr 19, 2005, at 1:01 AM, Tom Lane wrote:

> Lorenzo Thurman <lorenzo@thethurmans.com> writes:
>> I'm trying that right now. I think there may be mis-match in the build
>> settings between upgrades of postgresql. The "USE" settings may be at
>> fault:
>
>>   - - pg-hier        : Enables recursive queries like Oracle's
>> 'CONNECT
>> BY' feature.
>
> [ rolls eyes... ]  Yup, that's Gentoo all right: throw in random
> patches
> that have been rejected by the upstream developers.  Now that I think
> about it, this failure is exactly what that patch is known to cause,
> because it makes an incompatible change in Query structures and hence
> in on-disk view rule representation.
>
>> I think these may have been changed since the original install.
>
> Go back to your prior setting, or even better stop using Gentoo's
> hacked-up version.  I'm not sure why we even bother to answer support
> requests from Gentoo users, when what they are using is not our
> software but some randomly-modified variant.  I wonder what other
> brokennesses Gentoo may be including ...
>
> (Just for the record: I work for Red Hat, which has a rather different
> notion of the level of reliability it wants to ship.  So take my
> opinion
> with the appropriate grain of salt.  But I'm a mite ticked off at the
> moment --- you're not the first person to have been bitten by this,
> and you likely won't be the last, and I think it's entirely because
> Gentoo has such a low quality standard for the patches they ship.)
>
>             regards, tom lane
>
>


Tech/Library Combo Lab Manager
Northwestern University
Office Tech MG49
mailto:lthurman@northwestern.edu
voice: 847-467-6565
pager: 847-536-0094

On Tue, 19 Apr 2005 11:53 pm, Lorenzo Thurman wrote:
> Thanks for the reply. I've tried recompiling with my install build
> settings, but no luck. I've posted a message on the Gentoo forums.
> Hopefully they will have an answer. If they do, I'll post back here for
> future reference.
>
I read your post in the forums.  And as Tom suggested, it's going nothing to do with pg_dump,
you need to remerge postgresql at the very least, and with some C and USE flags you understand.

The Usual Gentoo causes come to mind first.  USE flags set correctly?  what are they?

What are your GCC flags.  I see a lot of gentoo users who just about turn on every compiler flag
without actually knowing what they do, or how they effect things.  Are your C_FLAGS conservative?

I've been using Postgresql on gentoo for both 7.4, and 8.0 from beta to 8.0.2 with no problems.  But then
I always set my C_FLAGS to something conservative like CGLAGS="-march=i586 -mcpu=i586 -O2 -pipe"
yes, it may seems a "Gentoo Conservative" buy I don't get broken software.  Always check extra patches
applied to the default distribution if you ever have trouble to weed out problem.  And never build with and
USE flags you don't understand the implications of.  Especially package specific ones.

I understand Tom's frustration, as Redhat is in business and ships quality checked software, and Gentoo
is run by a community group.  Of which I think may of the packagers are not tied to the projects they are
packaging.  But I also think there is often fault with the Gentoo user attempting to bleed his system a little
too much for speed, without considering the stability or even understand it.

Regards

Russell Smith.

> On Apr 19, 2005, at 1:01 AM, Tom Lane wrote:
>
> > Lorenzo Thurman <lorenzo@thethurmans.com> writes:
> >> I'm trying that right now. I think there may be mis-match in the build
> >> settings between upgrades of postgresql. The "USE" settings may be at
> >> fault:
> >
> >>   - - pg-hier        : Enables recursive queries like Oracle's
> >> 'CONNECT
> >> BY' feature.
> >
> > [ rolls eyes... ]  Yup, that's Gentoo all right: throw in random
> > patches
> > that have been rejected by the upstream developers.  Now that I think
> > about it, this failure is exactly what that patch is known to cause,
> > because it makes an incompatible change in Query structures and hence
> > in on-disk view rule representation.
> >
> >> I think these may have been changed since the original install.
> >
> > Go back to your prior setting, or even better stop using Gentoo's
> > hacked-up version.  I'm not sure why we even bother to answer support
> > requests from Gentoo users, when what they are using is not our
> > software but some randomly-modified variant.  I wonder what other
> > brokennesses Gentoo may be including ...
> >
> > (Just for the record: I work for Red Hat, which has a rather different
> > notion of the level of reliability it wants to ship.  So take my
> > opinion
> > with the appropriate grain of salt.  But I'm a mite ticked off at the
> > moment --- you're not the first person to have been bitten by this,
> > and you likely won't be the last, and I think it's entirely because
> > Gentoo has such a low quality standard for the patches they ship.)
> >
> >    regards, tom lane
> >
> >
>
>
> Tech/Library Combo Lab Manager
> Northwestern University
> Office Tech MG49
> mailto:lthurman@northwestern.edu
> voice: 847-467-6565
> pager: 847-536-0094
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 6: Have you searched our list archives?
>
>                http://archives.postgresql.org
>
>

What I was trying to do was export the database on one computer and
import it onto another. I gave up trying to fix the export problem
since I had an old backup of the database. It was old enough that it
was short three tables, but I have the raw tab delimited data so I
just reconstructed the database on this new machine. I've been running
Gentoo for about a year and a half now, and in the early days, I did
not fully understand all of the possible USE settings, but I've gotten
more comfortable with it over time. As you probably know, once you've
settled on what your USE settings should be, you can rebuild your
system to reflect those new settings. I did that, and since everything
appeared to be working OK, I assumed everything was OK, but obviously
the damage to PostgreSQL was already done. Anyway, I think my settings
now are pretty conservative and I know ot to play around with the
Postgres USE flags. One of the reasons I'm migrating is to do a
complete rebuild and apply what I've learned about Gentoo from scratch
on a new computer. Here are my settings, as you asked. I don't think
they're too out of line, but...

On Apr 19, 2005, at 6:06 p, Russell Smith wrote:


<excerpt>I read your post in the forums.  And as Tom suggested, it's
going nothing to do with pg_dump,

you need to remerge postgresql at the very least, and with some C and
USE flags you understand.


The Usual Gentoo causes come to mind first.  USE flags set correctly?
what are they?

</excerpt>


<color><param>0000,0000,DDDD</param>USE="X -gnome -gtk -gtk2  cups
-kde -qt  </color>


<excerpt>What are your GCC flags.  I see a lot of gentoo users who
just about turn on every compiler flag

without actually knowing what they do, or how they effect things.  Are
your C_FLAGS conservative?


</excerpt>

CFLAGS="-O2 -mtune=G3 -fno-strict-aliasing -pipe"

<excerpt>I've been using Postgresql on gentoo for both 7.4, and 8.0
from beta to 8.0.2 with no problems.  But then

I always set my C_FLAGS to something conservative like
CGLAGS="-march=i586 -mcpu=i586 -O2 -pipe"

yes, it may seems a "Gentoo Conservative" buy I don't get broken
software.  Always check extra patches

applied to the default distribution if you ever have trouble to weed
out problem.  And never build with and

USE flags you don't understand the implications of.  Especially
package specific ones.


</excerpt>I've always been a bit concerned about the patches myself.

<excerpt>I understand Tom's frustration, as Redhat is in business and
ships quality checked software, and Gentoo

is run by a community group.  Of which I think may of the packagers
are not tied to the projects they are

packaging.  But I also think there is often fault with the Gentoo user
attempting to bleed his system a little

too much for speed, without considering the stability or even
understand it.


</excerpt><smaller>


"My Break-Dancing days are over, but there's always the Funky Chicken"

--The Full Monty</smaller>

What I was trying to do was export the database on one computer and
import it onto another. I gave up trying to fix the export problem
since I had an old backup of the database. It was old enough that it
was short three tables, but I have the raw tab delimited data so I just
reconstructed the database on this new machine. I've been running
Gentoo for about a year and a half now, and in the early days, I did
not fully understand all of the possible USE settings, but I've gotten
more comfortable with it over time. As you probably know, once you've
settled on what your USE settings should be, you can rebuild your
system to reflect those new settings. I did that, and since everything
appeared to be working OK, I assumed everything was OK, but obviously
the damage to PostgreSQL was already done. Anyway, I think my settings
now are pretty conservative and I know ot to play around with the
Postgres USE flags. One of the reasons I'm migrating is to do a
complete rebuild and apply what I've learned about Gentoo from scratch
on a new computer. Here are my settings, as you asked. I don't think
they're too out of line, but...
On Apr 19, 2005, at 6:06 p, Russell Smith wrote:

> I read your post in the forums.  And as Tom suggested, it's going
> nothing to do with pg_dump,
> you need to remerge postgresql at the very least, and with some C and
> USE flags you understand.
>
> The Usual Gentoo causes come to mind first.  USE flags set correctly?
> what are they?


USE="X -gnome -gtk -gtk2  cups -kde -qt

> What are your GCC flags.  I see a lot of gentoo users who just about
> turn on every compiler flag
> without actually knowing what they do, or how they effect things.  Are
> your C_FLAGS conservative?
>

CFLAGS="-O2 -mtune=G3 -fno-strict-aliasing -pipe"
> I've been using Postgresql on gentoo for both 7.4, and 8.0 from beta
> to 8.0.2 with no problems.  But then
> I always set my C_FLAGS to something conservative like
> CGLAGS="-march=i586 -mcpu=i586 -O2 -pipe"
> yes, it may seems a "Gentoo Conservative" buy I don't get broken
> software.  Always check extra patches
> applied to the default distribution if you ever have trouble to weed
> out problem.  And never build with and
> USE flags you don't understand the implications of.  Especially
> package specific ones.
>
I've always been a bit concerned about the patches myself.
> I understand Tom's frustration, as Redhat is in business and ships
> quality checked software, and Gentoo
> is run by a community group.  Of which I think may of the packagers
> are not tied to the projects they are
> packaging.  But I also think there is often fault with the Gentoo user
> attempting to bleed his system a little
> too much for speed, without considering the stability or even
> understand it.
>


"My Break-Dancing days are over, but there's always the Funky Chicken"
--The Full Monty