Re: [PATCH 0/1] Update _PC_ASYNC_IO return value

[Corinna Vinschen <corinna-cygwin at cygwin dot com>]

On Jul 26 12:17, Corinna Vinschen wrote:
> On Jul 26 01:51, Mark Geisert wrote:
> > Hi Corinna,
> > 
> > On Thu, 26 Jul 2018, Corinna Vinschen wrote:
> > > On Jul 25 15:06, Yaakov Selkowitz wrote:
> > > > > From discussion on IRC:
> > > > 
> > > > <yselkowitz> corinna, just sent a patch for _POSIX_ASYNCHRONOUS_IO as a
> > > > 	  follow-up to the AIO feature, but am still wondering about
> > > > 	  _[POSIX|PC]_ASYNC_IO
> > > > [snip]
> > > > <corinna> in terms of _PC_ASYNC_IO, the test might be a bit tricky
> > > > <corinna> let me check
> > > > <corinna> actually, no
> > > > <corinna> it's easy
> > > > <corinna> Mark implemented the stuff with pread/pwrite only on disk files
> > > > <corinna> but otherwise it's device independent in that he implemented a
> > > > 	  workaround for pipes and stuff
> > > > <corinna> so, in theory we can just return 1
> > > > 
> > > > I'm not sure how to test this atm, but based on the above I have made
> > > > the following patch so this doesn't get lost.
> > > > 
> > > > Yaakov Selkowitz (1):
> > > >   Cygwin: fpathconf: update _PC_ASYNC_IO return value
> > > > 
> > > >  winsup/cygwin/fhandler.cc | 1 +
> > > >  1 file changed, 1 insertion(+)
> > > > 
> > > > --
> > > > 2.17.1
> > > 
> > > Mark?  Any comment you want to make?
> > 
> > Thanks for asking.  Your characterization of my implementation is correct.
> > The intent is for aio_* async I/O to be supported on all descriptors.  On
> > the most useful case of binary local disk files, inline pread|pwrite is
> > used.  But I wanted to make sure the AIO interface would do the right thing
> > on other kinds of descriptors without bothering the user about it.
> > 
> > So if the intent of the _PC_ASYNC_IO flag is to say that async I/O is
> > supported generally, I do think setting it to 1 is appropriate.  That is,
> > if it's talking about the aio_* interfaces.  If there's an O_ASYNC defined
> > for app coders, my recent contribution doesn't address that at all.
> 
> Good question.  O_ASYNC is a BSD invention, and it's not defined in
> POSIX at all.  Since we're in POSIX territory, _PC_ASYNC_IO and
> _POSIX_ASYNC_IO can only refer to async io as implemented by your new
> aio code:
> 
> - _POSIX_ASYNCHRONOUS_IO defines if the implementation supports async io
>   at all.
> - _POSIX_ASYNC_IO defines if the file in question supports async io.

Offline all day due to a major hardware breakdown.  The heat or something.
My server died the mainboard death.  I'm running a replacement system
right now with *much* less power.  Will take a week or more to get an
adequate replacement so I'm on half power for a while.

Oh well, enough lamenting.

Yaakov, please push your patch.


Thanks,
Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

Attachment: signature.asc
Description: PGP signature