MIME-Version: 1.0
References: <3f22a8bb-29e8-40cc-97a1-309181da2c13@dunslane.net>
 <CAC6VRoYWPQZeLDuKVNeMz5z-iiVnv0zWyBwWhxdm4UQO17-AUQ@mail.gmail.com>
 <CAKYtNAoNi8O1wbSZx0a1oSi-YzQtprxqpfZqryRq39hXn_p-OA@mail.gmail.com>
 <CAC6VRoYCqJq=0q3GF+vfDVbY4JUehqsuxYhcmVi7A=1hfNzSJg@mail.gmail.com>
 <CAKYtNApm8nA6hVMNkSjeNPiwTZBWYkhJKU7haQghUZdw74tYxA@mail.gmail.com>
 <CACJufxHc5Rp2roGyJOVnCk=Mx8_42_9vCGJSCFpLB5=AFQqHTQ@mail.gmail.com>
 <CAKYtNAoz-Sz2X1ipkQ5x1Sh08AMyy1kDRGfeRW3P1W-QoN+WOg@mail.gmail.com>
 <CAC6VRoZU2L2GX7+bXpsZ6bo+NuxBOe1dTGDPzKobUL4KVdm46A@mail.gmail.com>
 <CAKYtNApq+6Riz26So1nmXkRpL_7o868+r6i=wxQ+AgGQ4oSLig@mail.gmail.com>
 <CAC6VRoZbP=-=a+a78RoWcs2L4=4VEDZqJMg+7SyUAcSHATyDAQ@mail.gmail.com>
 <CAC6VRoY-1WC_O5UDwKzNc18LXLwjuTFb+qMNY3gaF3cCov6M6g@mail.gmail.com>
 <CAKYtNArc3j8Akb+nB1QyJLKic7ej7vKZqKnv_mbB+iQAJRzLnA@mail.gmail.com>
 <CAC6VRoat6NPcS6vfGrPrwaQB3G1JHLSqDeAkf1AJ7npWrwhnUQ@mail.gmail.com>
 <CAKYtNAp=TFEZWQbTRZso3kNaD9BH=-JPNQ30mT60cJN_04hEog@mail.gmail.com> <4022765f-38ee-48a3-b246-615b3f8e1c23@dunslane.net>
In-Reply-To: <4022765f-38ee-48a3-b246-615b3f8e1c23@dunslane.net>
From: Mahendra Singh Thalor <mahi6run@gmail.com>
Date: Fri, 20 Feb 2026 19:40:36 +0530
Message-ID: <CAKYtNAq73_UeL4-0wgu9rawctHdS2zhpzWtqVuOChCS3M+2Ymg@mail.gmail.com>
Subject: Re: Non-text mode for pg_dumpall
To: Andrew Dunstan <andrew@dunslane.net>
Cc: tushar <tushar.ahuja@enterprisedb.com>, jian he <jian.universality@gmail.com>, 
	Vaibhav Dalvi <vaibhav.dalvi@enterprisedb.com>, pgsql-hackers@lists.postgresql.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: <https://www.postgresql.org/message-id/CAKYtNAq73_UeL4-0wgu9rawctHdS2zhpzWtqVuOChCS3M%2B2Ymg%40mail.gmail.com>
Precedence: bulk

On Fri, 20 Feb 2026 at 01:30, Andrew Dunstan <andrew@dunslane.net> wrote:
>
>
> On 2026-02-18 We 12:15 AM, Mahendra Singh Thalor wrote:
>
> On Wed, 28 Jan 2026 at 13:04, tushar <tushar.ahuja@enterprisedb.com> wrot=
e:
> >
> >
> >
> > On Tue, Jan 27, 2026 at 9:11=E2=80=AFPM Mahendra Singh Thalor <mahi6run=
@gmail.com> wrote:
> >>
> >> On Fri, 23 Jan 2026 at 19:07, tushar <tushar.ahuja@enterprisedb.com> w=
rote:
> >> >
> >> >
> >> >
> >> > On Fri, Jan 23, 2026 at 12:21=E2=80=AFPM tushar <tushar.ahuja@enterp=
risedb.com> wrote:
> >> >>
> >> >>
> >> >> Thanks Mahendra, a minor  observation -  The pg_restore output show=
s a double slash in the map.dat path (e.g., abc.tar//map.dat).
> >> >> While it doesn't break the restore, we may want to clean up the pat=
h joining logic.
> >> >>
> >> >> [edb@1a1c15437e7c bin]$ ./pg_restore -Ft -C abc.tar/ -d postgres -p=
 9011  -U  ed -v
> >> >> pg_restore: found database "template1
> >> >> " (OID: 1) in file "abc.tar//map.dat"
> >> >> pg_restore: found database "postgres
> >> >> " (OID: 5) in file "abc.tar//map.dat"
> >> >>
> >> >>
> >> >
> >> > Please refer to this scenario where - Objects  created under templat=
e1 and the postgres database by a specific user are failing during a cross-=
cluster restore.
> >> > When restoring to a new cluster as a different superuser, pg_restore=
 throws the error: ERROR: role "edb" does not exist.
> >> > It appears the restore is attempting to preserve the original owners=
hip of template1 objects even when the target environment lacks those speci=
fic roles.
> >> >
> >> > Steps to reproduce:
> >> > initdb ( ./initdb -U edb -D data) , start the server , connect to po=
stgres and  template1 database one by one and  create
> >> > this table ( create table test(n int); )
> >> > perform pg_dumpall operation ( ./pg_dumpall -Ft -f abc.tar)
> >> > initdb (./initdb -U xyz) , start the server , create a database ( cr=
eate database abc;)
> >> > perform pg_restore operation ( ./pg_restore -Ft -C abc.tar/ -d postg=
res -p 9033 -U xyz)
> >> > --getting an error,  table 'test' will be created on 'template1' dat=
abase but failed to create on an another database ( in this case - 'abc' da=
tabase)
> >> >
> >> > regards,
> >>
> >> Hi,
> >> Here I am attaching an updated patch for the review and testing.
> >> Thanks Jian for the reporting rebase issue.
> >>
> >
> > Thanks Mahendra, getting a regression error during the restore process =
after applying this patch.
> >
> >  [edb@1a1c15437e7c bin]$ ./pg_restore -Ft -C abc1.tar/ -d postgres -p 9=
000
> > pg_restore: error: could not execute query: ERROR:  non-standard string=
 literals are not supported
> > Command was: SET standard_conforming_strings =3D off;
> > pg_restore: warning: errors ignored on restore: 1
> >
> > in earlier patches - this was not coming.
> >
> > regards,
> >
>
> Thanks Andrew for some design related feedback.
>
> Thanks Jian for the offline discussions, reviews, testing and delta patch=
es.
>
> Thanks Tushar for the detailed testing.
>
> Brief about this patch:
> new option to pg_dumpall: --format=3Dd/t/c/p directory/tar/custam/plain
>
> If the user gives a non-text format with pg_dumpall command, then the ful=
l cluster will be dumped and global objects (roles. tablespaces, databases)=
 will be dumped into toc.glo file in custom format with drop commands and d=
atabases will be dumped into a given archive format one by one with oid.tar=
/oid.dmp/oid files/dir.
> When restoring, if the user gives -g(globals-only) option, then creating =
commands of  only global users/tablespaces/databases will be restored. (no =
drop commands will be executed)
> toc.glo will be executed with -e(exit-on-error=3Dfalse) and --transaction=
-size=3D0 as some user already created. If the user wants to restore a sing=
le database, they can restore it by a single dump file. For --clean and -g(=
globals-only), we added some error cases so that roles/databases/tablespace=
s will not be dropped.
>
> Here, I am attaching an updated patch for the review and testing.
>
>
> Here's an update after a round of review. Most of the changes are pretty =
minor, but it should get the cfbot all green, with a Windows fix in the tes=
ts.
>
>
> cheers
>
>
> andrew
>
>
> --
> Andrew Dunstan
> EDB: https://www.enterprisedb.com

Thanks Andrew for the updated patch.

Updated patch looks good to me. I have verified by some dump/restore testin=
g.

--=20
Thanks and Regards
Mahendra Singh Thalor
EnterpriseDB: http://www.enterprisedb.com