Adding aio.h and mqueue.h

Adding aio.h and mqueue.h

[Joel Sherrill]

[-- Attachment #1: Type: text/plain, Size: 389 bytes --]

Hi

We have migrated almost all of our POSIX defined header files from RTEMS
itself to newlib.  I think aio.h and mqueue.h are the last POSIX headers
left on our side.

What's the feeling on merging these? Would it be ok to put them in
newlib/libc/include?

Our version of the headers have Doxygen comments. Would these be OK in
newlib or do I need to remove the Doxygen?

Thanks.

--joel

Re: Adding aio.h and mqueue.h

[Jeff Johnston]

[-- Attachment #1: Type: text/plain, Size: 583 bytes --]

As long as the licensing is shared-newlib-compatible (non-GPL), it should
be ok.

-- Jeff J.

On Thu, Dec 1, 2022 at 11:37 AM Joel Sherrill <joel@rtems.org> wrote:

> Hi
>
> We have migrated almost all of our POSIX defined header files from RTEMS
> itself to newlib.  I think aio.h and mqueue.h are the last POSIX headers
> left on our side.
>
> What's the feeling on merging these? Would it be ok to put them in
> newlib/libc/include?
>
> Our version of the headers have Doxygen comments. Would these be OK in
> newlib or do I need to remove the Doxygen?
>
> Thanks.
>
> --joel
>
>

Re: Adding aio.h and mqueue.h

[Corinna Vinschen]

On Dec  1 18:22, Jeff Johnston wrote:
> As long as the licensing is shared-newlib-compatible (non-GPL), it should
> be ok.
> 
> -- Jeff J.

Cygwin already comes with aio.h and mqueue.h headers.

Ideally they are reused for newlib, or the new aio.h/mqueue.h headers
are checked that they provide the same definitions and replace the
Cygwin-only ones.


Thanks,
Corinna





> 
> On Thu, Dec 1, 2022 at 11:37 AM Joel Sherrill <joel@rtems.org> wrote:
> 
> > Hi
> >
> > We have migrated almost all of our POSIX defined header files from RTEMS
> > itself to newlib.  I think aio.h and mqueue.h are the last POSIX headers
> > left on our side.
> >
> > What's the feeling on merging these? Would it be ok to put them in
> > newlib/libc/include?
> >
> > Our version of the headers have Doxygen comments. Would these be OK in
> > newlib or do I need to remove the Doxygen?
> >
> > Thanks.
> >
> > --joel
> >
> >

Re: Adding aio.h and mqueue.h

[Joel Sherrill]

[-- Attachment #1: Type: text/plain, Size: 1481 bytes --]

On Fri, Dec 2, 2022 at 4:07 AM Corinna Vinschen <vinschen@redhat.com> wrote:

> On Dec  1 18:22, Jeff Johnston wrote:
> > As long as the licensing is shared-newlib-compatible (non-GPL), it should
> > be ok.
> >
> > -- Jeff J.
>
> Cygwin already comes with aio.h and mqueue.h headers.
>
> Ideally they are reused for newlib, or the new aio.h/mqueue.h headers
> are checked that they provide the same definitions and replace the
> Cygwin-only ones.
>

OK. So the same defines with the same values? Anything cygwin specific
goes into an ifdef, etc. I vaguely recall doing this before for other
header files.

RTEMS is always built from source so changing the values isn't a huge
binary compatibility issue. But moving the header file from RTEMS to
newlib will require everyone using the git master to build new tools.
So there is a bit of coordination on our side.

--joel

>
>
> Thanks,
> Corinna
>
>
>
>
>
> >
> > On Thu, Dec 1, 2022 at 11:37 AM Joel Sherrill <joel@rtems.org> wrote:
> >
> > > Hi
> > >
> > > We have migrated almost all of our POSIX defined header files from
> RTEMS
> > > itself to newlib.  I think aio.h and mqueue.h are the last POSIX
> headers
> > > left on our side.
> > >
> > > What's the feeling on merging these? Would it be ok to put them in
> > > newlib/libc/include?
> > >
> > > Our version of the headers have Doxygen comments. Would these be OK in
> > > newlib or do I need to remove the Doxygen?
> > >
> > > Thanks.
> > >
> > > --joel
> > >
> > >
>
>

Re: Adding aio.h and mqueue.h

[Corinna Vinschen]

On Dec  2 08:15, Joel Sherrill wrote:
> On Fri, Dec 2, 2022 at 4:07 AM Corinna Vinschen <vinschen@redhat.com> wrote:
> 
> > On Dec  1 18:22, Jeff Johnston wrote:
> > > As long as the licensing is shared-newlib-compatible (non-GPL), it should
> > > be ok.
> > >
> > > -- Jeff J.
> >
> > Cygwin already comes with aio.h and mqueue.h headers.
> >
> > Ideally they are reused for newlib, or the new aio.h/mqueue.h headers
> > are checked that they provide the same definitions and replace the
> > Cygwin-only ones.
> >
> 
> OK. So the same defines with the same values? Anything cygwin specific
> goes into an ifdef, etc. I vaguely recall doing this before for other
> header files.

There shouldn't be much Cygwin-specifc in these files and they are quite
short.  Please check the Cygwin files against the files from RTEMS.
Maybe it makes sense to use them instead and just add the (minor)
differences to Cygwin.  If the types used have the same size and the
same signedness, there shouldn't be much of a problem anyway.

However, I'd like to defer this by a few days or weeks.  We're just
in the beta phase for Cygwin 3.4, and I would rather have stable
headers for now :}


Thanks,
Corinna

Re: Adding aio.h and mqueue.h

[Joel Sherrill]

[-- Attachment #1: Type: text/plain, Size: 1553 bytes --]

On Fri, Dec 2, 2022, 11:10 AM Corinna Vinschen <vinschen@redhat.com> wrote:

> On Dec  2 08:15, Joel Sherrill wrote:
> > On Fri, Dec 2, 2022 at 4:07 AM Corinna Vinschen <vinschen@redhat.com>
> wrote:
> >
> > > On Dec  1 18:22, Jeff Johnston wrote:
> > > > As long as the licensing is shared-newlib-compatible (non-GPL), it
> should
> > > > be ok.
> > > >
> > > > -- Jeff J.
> > >
> > > Cygwin already comes with aio.h and mqueue.h headers.
> > >
> > > Ideally they are reused for newlib, or the new aio.h/mqueue.h headers
> > > are checked that they provide the same definitions and replace the
> > > Cygwin-only ones.
> > >
> >
> > OK. So the same defines with the same values? Anything cygwin specific
> > goes into an ifdef, etc. I vaguely recall doing this before for other
> > header files.
>
> There shouldn't be much Cygwin-specifc in these files and they are quite
> short.  Please check the Cygwin files against the files from RTEMS.
> Maybe it makes sense to use them instead and just add the (minor)
> differences to Cygwin.  If the types used have the same size and the
> same signedness, there shouldn't be much of a problem anyway.
>
> However, I'd like to defer this by a few days or weeks.  We're just
> in the beta phase for Cygwin 3.4, and I would rather have stable
> headers for now :
>

We are also trying to get ready for a release branch. I think I'll push our
tickets to a new milestone. That seems to be better for both projects.

With any luck, I can revisit the long double code around then also.

>
>
> Thanks,
> Corinna
>
>

Re: Adding aio.h and mqueue.h

[Corinna Vinschen]

On Dec  2 11:19, Joel Sherrill wrote:
> On Fri, Dec 2, 2022, 11:10 AM Corinna Vinschen <vinschen@redhat.com> wrote:
> 
> > On Dec  2 08:15, Joel Sherrill wrote:
> > > On Fri, Dec 2, 2022 at 4:07 AM Corinna Vinschen <vinschen@redhat.com>
> > wrote:
> > >
> > > > On Dec  1 18:22, Jeff Johnston wrote:
> > > > > As long as the licensing is shared-newlib-compatible (non-GPL), it
> > should
> > > > > be ok.
> > > > >
> > > > > -- Jeff J.
> > > >
> > > > Cygwin already comes with aio.h and mqueue.h headers.
> > > >
> > > > Ideally they are reused for newlib, or the new aio.h/mqueue.h headers
> > > > are checked that they provide the same definitions and replace the
> > > > Cygwin-only ones.
> > > >
> > >
> > > OK. So the same defines with the same values? Anything cygwin specific
> > > goes into an ifdef, etc. I vaguely recall doing this before for other
> > > header files.
> >
> > There shouldn't be much Cygwin-specifc in these files and they are quite
> > short.  Please check the Cygwin files against the files from RTEMS.
> > Maybe it makes sense to use them instead and just add the (minor)
> > differences to Cygwin.  If the types used have the same size and the
> > same signedness, there shouldn't be much of a problem anyway.
> >
> > However, I'd like to defer this by a few days or weeks.  We're just
> > in the beta phase for Cygwin 3.4, and I would rather have stable
> > headers for now :
> >
> 
> We are also trying to get ready for a release branch. I think I'll push our
> tickets to a new milestone. That seems to be better for both projects.
> 
> With any luck, I can revisit the long double code around then also.

I released Cygwin 3.4.0 today.  Further 3.4 bugfixing will take place
on a branch, so the master branch is open to more intrusive changes now.


Corinna