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

On Mon, Oct 31, 2022 at 5:01 AM Michael Paquier <michael(at)paquier(dot)xyz> wrote:
>
> On Sun, Oct 30, 2022 at 03:44:32PM +0100, Alvaro Herrera wrote:
> > So I'm kinda proposing that we only do the forward struct initialization
> > dance when it really saves on things -- in particular, when it helps
> > avoid or reduce massive indirect header inclusion.
>
> Sure.

I don't think including pg_iovec.h in file_utils.h is a good idea. I
agree that pg_iovec.h is fairly a small header file but file_utils.h
is included in 21 .c files, as of today and the file_utils.h footprint
might increase in future. Therefore, I'd vote for forward struct
initialization as it is on HEAD today.

> > extern ssize_t pg_pwritev_with_retry(int fd,
> > - const struct iovec *iov,
> > + const iovec *iov,
> > int iovcnt,
> > off_t offset);
>
> However this still needs to be defined as a struct, no?

Yes, we need a struct there because we haven't typedef'ed struct iovec.

Also, the patch forgets to remove "port/pg_iovec.h" from file_utils.c

--
Bharath Rupireddy
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com