Re: current_logfiles not following group access and instead followslog_file_mode permissions - Mailing list pgsql-hackers

From Haribabu Kommi
Subject Re: current_logfiles not following group access and instead followslog_file_mode permissions
Date
Msg-id CAJrrPGfAjK8sFYknpNjor800Thi12XeyPUNirvwYP_tCyK=16Q@mail.gmail.com
Whole thread Raw
In response to Re: current_logfiles not following group access and instead followslog_file_mode permissions  (Michael Paquier <michael@paquier.xyz>)
List pgsql-hackers

On Sun, Mar 24, 2019 at 11:16 PM Michael Paquier <michael@paquier.xyz> wrote:
On Fri, Mar 22, 2019 at 01:01:44PM +0900, Michael Paquier wrote:
> On Fri, Mar 22, 2019 at 02:35:41PM +1100, Haribabu Kommi wrote:
> > Thanks for the correction.  Yes, that is correct and it works fine.
>
> Thanks for double-checking.  Are there any objections with this patch?

Done and committed down to v11 where group access has been added.
There could be an argument to do the same in v10, but as the root of
the problem is the interaction between a data folder using 0640 as
base mode for files and log_file_mode being more restrictive, then it
cannot apply to v10.

After testing and reviewing the patch, I noticed that all versions
sent up to now missed two things done by logfile_open():
- Bufferring is line-buffered.  For current_logfiles it may not matter
much as the contents are first written into a temporary file and then
the file is renamed, but for debugging having the insurance of
consistent contents is nice even for the temporary file.
- current_logfiles uses \r\n.  While it does not have a consequence
for the parsing of the file by pg_current_logfile, it breaks the
readability of the file on Windows, which is not nice.
So I have kept the patch with the previous defaults for consistency.
Perhaps they could be changed, but the current set is a good set.

Thanks Micheal and others.
This really helps to choose the restrictive log file permissions even when
the data directory is enabled with group access.

Regards,
Haribabu Kommi
Fujitsu Australia

pgsql-hackers by date:

Previous
From: Michael Paquier
Date:
Subject: Re: reorder pg_rewind control file sync
Next
From: David Steele
Date:
Subject: Re: Re: make installcheck-world in a clean environment