Checking pgwin32_is_junction() errors

Thomas Munro <thomas.munro@gmail.com>

From: Thomas Munro <thomas.munro@gmail.com>

To: pgsql-hackers <pgsql-hackers@postgresql.org>

Date: 2022-03-24T03:30:26Z

Lists: pgsql-hackers

Attachments

0001-Raise-errors-in-pgwin32_is_junction.patch (text/x-patch) patch 0001

Hi,

The comment for pgwin32_is_junction() says "Assumes the file exists,
so will return false if it doesn't (since a nonexistent file is not a
junction)".  In fact that's the behaviour for any kind of error, and
although we set errno in that case, no caller ever checks it.

I think it'd be better to add missing_ok and elevel parameters,
following existing patterns.  Unfortunately, it can't use the generic
frontend logging to implement elevel in frontend code from its current
location, because pgport can't call pgcommon.  For now I came up with
a kludge to work around that problem, but I don't like it, and would
need to come up with something better...

Sketch code attached.

Commits

Fix readlink() for non-PostgreSQL junction points on Windows.
- f71007fbb3b8 16.0 landed
Fix stat() for recursive junction points on Windows.
- 4517358ee782 16.0 landed
aix: fix misreading of condition in 8f12a4e7add
- 7e29a79a46d3 16.0 cited
Make unlink() work for junction points on Windows.
- f357233c9db8 16.0 landed
Replace pgwin32_is_junction() with lstat().
- 5fc88c5d53e4 16.0 landed
Provide lstat() for Windows.
- c5cb8f3b770c 16.0 landed