On Tue, Jul 20, 2021 at 12:26 PM Tom Lane wrote:
> > I can try that on the gcc farm in a bit.
Thanks!
> Hmm, it compiles cleanly, but something seems drastically wrong,
> because performance is just awful. On the other hand, I don't
> know what sort of storage is underlying this instance, so
I wrote:
> Thomas Munro writes:
>> While I was here again, I couldn't resist trying to extend this to
>> Solaris, since it looked so easy. I don't have access, but I tested
>> on Illumos by undefining O_DIRECT. Thoughts?
> I can try that on the gcc farm in a bit.
Hmm, it compiles cleanly, but
Thomas Munro writes:
> While I was here again, I couldn't resist trying to extend this to
> Solaris, since it looked so easy. I don't have access, but I tested
> on Illumos by undefining O_DIRECT. Thoughts?
I can try that on the gcc farm in a bit.
regards, tom lane
On Tue, Jul 20, 2021 at 2:13 AM Tom Lane wrote:
> Hmm ... we used to have to avoid putting #if constructs in the arguments
> of macros (such as StaticAssertStmt). Maybe that's not a thing anymore
> with C99, and in any case this whole stanza is fairly platform-specific
> so we may not run into a
Thomas Munro writes:
> On Mon, Jul 19, 2021 at 4:42 PM Tom Lane wrote:
>> prairiedog thinks that Assert is too optimistic about whether all
>> those flags exist.
> Fixed.
Hmm ... we used to have to avoid putting #if constructs in the arguments
of macros (such as StaticAssertStmt). Maybe
On Mon, Jul 19, 2021 at 12:55 AM Thomas Munro
wrote:
>
> On Mon, Jul 19, 2021 at 4:42 PM Tom Lane wrote:
> > prairiedog thinks that Assert is too optimistic about whether all
> > those flags exist.
>
> Fixed.
>
> (Huh, I received no -committers email for 2dbe8905.)
It didn't show up in the
On Mon, Jul 19, 2021 at 4:42 PM Tom Lane wrote:
> prairiedog thinks that Assert is too optimistic about whether all
> those flags exist.
Fixed.
(Huh, I received no -committers email for 2dbe8905.)
Thomas Munro writes:
> Agreed. Pushed!
prairiedog thinks that Assert is too optimistic about whether all
those flags exist.
regards, tom lane
On Tue, Jul 13, 2021 at 1:56 PM Andres Freund wrote:
> On 2021-07-13 13:25:50 +1200, Thomas Munro wrote:
> > I'm planning to go with that idea (#1), if there are no objections.
>
> The only other viable approach I see is to completely separate our
> internal flag representation from the OS
Hi,
On 2021-07-13 13:25:50 +1200, Thomas Munro wrote:
> On Mon, May 31, 2021 at 10:29 AM Thomas Munro wrote:
> > For comparison, here is my sketch of idea #1. I pick an arbitrary
> > value to use as PG_O_DIRECT (I don't want to define O_DIRECT for fear
> > of breaking other code that might see
On Mon, May 31, 2021 at 10:29 AM Thomas Munro wrote:
> For comparison, here is my sketch of idea #1. I pick an arbitrary
> value to use as PG_O_DIRECT (I don't want to define O_DIRECT for fear
> of breaking other code that might see it and try to pass it into
> open()... for all I know, it might
On Mon, May 31, 2021 at 4:19 AM Justin Pryzby wrote:
> Should there be an "else" to warn/error in the case that "direct" is requested
> but not supported?
The way we use O_DIRECT currently is extremely minimal, it's just "if
you've got it, we'll use it, but otherwise not complain", and I wasn't
On Mon, May 31, 2021 at 8:12 AM Andres Freund wrote:
> On 2021-05-30 16:39:48 +1200, Thomas Munro wrote:
> > I thought about a few different ways to encapsulate this API
> > difference in PostgreSQL, and toyed with two:
> >
> > 1. We could define our own fake O_DIRECT flag, and translate that to
Hi,
Thanks for starting the discussion on this!
On 2021-05-30 16:39:48 +1200, Thomas Munro wrote:
> I thought about a few different ways to encapsulate this API
> difference in PostgreSQL, and toyed with two:
>
> 1. We could define our own fake O_DIRECT flag, and translate that to
> the right
On Sun, May 30, 2021 at 04:39:48PM +1200, Thomas Munro wrote:
> +BasicOpenFilePermDirect(const char *fileName, int fileFlags, mode_t fileMode,
> + bool direct)
> ...
> +#if !defined(O_DIRECT) && defined(F_NOCACHE)
> + /*
Hi,
IRIX gave the world O_DIRECT, and then every Unix I've used followed
their lead except Apple's, which gave the world fcntl(fd, F_NOCACHE,
1). From what I could find in public discussion, this API difference
may stem from the caching policy being controlled at the per-file
(vnode) level in
16 matches
Mail list logo
