Horton

Re: Loaded footgun open_datasync on Windows

Michael Paquier <michael@paquier.xyz>

View thread View on postgr.es
From: Michael Paquier <michael@paquier.xyz>
To: Laurenz Albe <laurenz.albe@cybertec.at>
Cc: Noah Misch <noah@leadboat.com>, Kuntal Ghosh <kuntalghosh.2007@gmail.com>, Amit Kapila <amit.kapila16@gmail.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2018-09-14T07:31:47Z
Lists: pgsql-hackers 
On Fri, Sep 14, 2018 at 08:43:18AM +0200, Laurenz Albe wrote:
> Thanks for being interested and doing the work.

No problem.  I have a sort of Windows-label stuck on me for ages, and
those random buildfarm failures are annoying with TAP tests on Windows.

> If it turns out not to break anything, would you consider backpatching?
> On the one hand it fixes a bug, on the other hand it affects all
> frontend executables...

Yeah, for this reason I would not do a backpatch.  I have a very hard
time to believe that any frontend tools on Windows developed by anybody
rely on files to be opened only by a single process, still if they do
they would be surprised to see a change of behavior after a minor
update in case they rely on the concurrency limitations.

> I wonder why nobody noticed the problem in pg_test_fsync earlier.
> Is it that people running Windows care less if their storage is
> reliable?

likely so.
--
Michael

Commits

  1. Allow concurrent-safe open() and fopen() in frontend code for Windows

  2. Fix inclusions of c.h from .h files.

  3. Allow borland compiles.

  4. Corrects issues recently posted by Dann Corbit, allowing libpq/psql to