lowering pg_regress privileges on Windows
From the "scratch a long running itch" department.
The attached ridiculously tiny patch solves the problem whereby while we
can run Postgres on Windows safely from an Administrator account, we
can't run run the regression tests from the same account, since it fails
on the tablespace test, the tablespace directory having been set up
without first having lowered privileges. The solution is to lower
pg_regress' privileges in the same way that we do with other binaries.
This is useful in setups like Appveyor where running under any other
account is ... difficult. For the cfbot Thomas has had to make the
script hack the schedule file to omit the tablespace test. This would
make that redundant.
I propose to backpatch this. It's close enough to a bug and the risk is
almost infinitely small.
cheers
andrew
--
Andrew Dunstan https://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
Attachments:
regress_restricted.patchtext/x-patch; name=regress_restricted.patchDownload+2-0
On Thu, Oct 18, 2018 at 08:31:11AM -0400, Andrew Dunstan wrote:
The attached ridiculously tiny patch solves the problem whereby while we can
run Postgres on Windows safely from an Administrator account, we can't run
run the regression tests from the same account, since it fails on the
tablespace test, the tablespace directory having been set up without first
having lowered privileges. The solution is to lower pg_regress' privileges
in the same way that we do with other binaries. This is useful in setups
like Appveyor where running under any other account is ... difficult. For
the cfbot Thomas has had to make the script hack the schedule file to omit
the tablespace test. This would make that redundant.I propose to backpatch this. It's close enough to a bug and the risk is
almost infinitely small.
+1. get_restricted_token() refactoring has been done down to
REL9_5_STABLE. With 9.4 and older you would need to copy again this
full routine into pg_regress.c, which is in my opinion not worth
worrying about.
--
Michael
On Fri, Oct 19, 2018 at 1:13 PM Michael Paquier <michael@paquier.xyz> wrote:
On Thu, Oct 18, 2018 at 08:31:11AM -0400, Andrew Dunstan wrote:
The attached ridiculously tiny patch solves the problem whereby while we can
run Postgres on Windows safely from an Administrator account, we can't run
run the regression tests from the same account, since it fails on the
tablespace test, the tablespace directory having been set up without first
having lowered privileges. The solution is to lower pg_regress' privileges
in the same way that we do with other binaries. This is useful in setups
like Appveyor where running under any other account is ... difficult. For
the cfbot Thomas has had to make the script hack the schedule file to omit
the tablespace test. This would make that redundant.I propose to backpatch this. It's close enough to a bug and the risk is
almost infinitely small.+1. get_restricted_token() refactoring has been done down to
REL9_5_STABLE. With 9.4 and older you would need to copy again this
full routine into pg_regress.c, which is in my opinion not worth
worrying about.
FWIW here is a successful Appveyor build including the full test
schedule (CI patch attached in case anyone is interested). Woohoo!
Thanks for figuring that out Andrew. I will be very happy to remove
that wart from my workflows.
https://ci.appveyor.com/project/macdice/postgres/builds/19626669
--
Thomas Munro
http://www.enterprisedb.com
Attachments:
0001-CI-configuration-for-Appveyor.patchapplication/octet-stream; name=0001-CI-configuration-for-Appveyor.patchDownload+81-1
On 10/18/2018 08:25 PM, Thomas Munro wrote:
On Fri, Oct 19, 2018 at 1:13 PM Michael Paquier <michael@paquier.xyz> wrote:
On Thu, Oct 18, 2018 at 08:31:11AM -0400, Andrew Dunstan wrote:
The attached ridiculously tiny patch solves the problem whereby while we can
run Postgres on Windows safely from an Administrator account, we can't run
run the regression tests from the same account, since it fails on the
tablespace test, the tablespace directory having been set up without first
having lowered privileges. The solution is to lower pg_regress' privileges
in the same way that we do with other binaries. This is useful in setups
like Appveyor where running under any other account is ... difficult. For
the cfbot Thomas has had to make the script hack the schedule file to omit
the tablespace test. This would make that redundant.I propose to backpatch this. It's close enough to a bug and the risk is
almost infinitely small.+1. get_restricted_token() refactoring has been done down to
REL9_5_STABLE. With 9.4 and older you would need to copy again this
full routine into pg_regress.c, which is in my opinion not worth
worrying about.FWIW here is a successful Appveyor build including the full test
schedule (CI patch attached in case anyone is interested). Woohoo!
Thanks for figuring that out Andrew. I will be very happy to remove
that wart from my workflows.https://ci.appveyor.com/project/macdice/postgres/builds/19626669
Excellent. I'll apply back to 9.5 as Michael suggests.
Having got past that hurdle I encountered another one in the same area.
pg_upgrade gives up its privileges and is then unable to write things
like log files and analyze scripts.
The attached patch cures the problem, but it doesn't seem like the best
cure. Maybe there is a more secure way to do it. Essentially it saves
out the ACLS for the current directory and its subdirectories and then
allows everyone to write to them, right before running pg_upgrade. When
pg_upgrade is done it restores the saved ACLs.
Maybe someone who understands more about how this all works can suggest
a less blunt force approach.
cheers
andrew
--
Andrew Dunstan https://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services