Re: Use pg_pwritev_with_retry() instead of write() in dir_open_for_write() to avoid partial writes?

On Mon, Mar 6, 2023 at 5:30 PM Michael Paquier <michael@paquier.xyz> wrote:
> On Mon, Feb 20, 2023 at 01:54:00PM +0530, Bharath Rupireddy wrote:
> > I ran some tests on my dev system [1] and I don't see much difference
> > between v3 and v4. So, +1 for v3 patch (+ argument order swap) from
> > Andres to keep the code simple and elegant.
>
> This thread has stalled for a couple of weeks, so I have gone back to
> it.  Testing on a tmpfs I am not seeing a difference if performance
> for any of the approaches discussed.  At the end, as I am the one
> behind the introduction of pg_pwrite_zeros(), I have applied v3 after
> switches the size and offset parameters to be the same way as in v4.

Apparently ye olde GCC 4.7 on "lapwing" doesn't like the way you
initialised that struct.  I guess it wants {{0}} instead of {0}.
Apparently old GCC was wrong about that warning[1], but that system
doesn't have the back-patched fixes?  Not sure.

[1] https://stackoverflow.com/questions/63355760/how-standard-is-the-0-initializer-in-c89