This is the mail archive of the
newlib@sourceware.org
mailing list for the newlib project.
RE: [AArch64] nosys.specs and printf
- From: Tamar Christina <Tamar dot Christina at arm dot com>
- To: Alexander Fedotov <alfedotov at gmail dot com>, Jon Beniston <jon at beniston dot com>
- Cc: "newlib at sourceware dot org" <newlib at sourceware dot org>, nd <nd at arm dot com>
- Date: Fri, 22 Sep 2017 08:05:13 +0000
- Subject: RE: [AArch64] nosys.specs and printf
- Authentication-results: sourceware.org; auth=none
- Authentication-results: spf=none (sender IP is ) smtp.mailfrom=Tamar dot Christina at arm dot com;
- Nodisclaimer: True
- References: <CAN8C2Cptf5v8+KrEsd7ppfjhbM88wYOianMcPVUJ19bm9SOBzA@mail.gmail.com> <DB6PR0802MB230915D25ABD3E6534040138FF600@DB6PR0802MB2309.eurprd08.prod.outlook.com> <CAN8C2Crygk-MsBz-=x5qa0pJ-2SvOQoOwdqcJztxppdTskEkdg@mail.gmail.com> <DB6PR0802MB23090588EC607014E69E9B1AFF600@DB6PR0802MB2309.eurprd08.prod.outlook.com> <CAN8C2CoX3aHHe2WixnQM1eHzigdLgb4A44v=BQj2Cs60tmmUHw@mail.gmail.com> <CAN8C2CraQaug7jdY=n3PuGroAJZtgM1vJgjRC=hFOQa8xy2knA@mail.gmail.com> <DB6PR0802MB2309953D1F436D074F15BA58FF610@DB6PR0802MB2309.eurprd08.prod.outlook.com> <CAN8C2Co+CR23ET---H5g71oxQ6BLTnqOR4hc4uZm=dxNJ7RRJw@mail.gmail.com> <023601d33219$e54c9950$afe5cbf0$@beniston.com> <CAN8C2Cp49aZ6cJ3TFZOeX_+7tLBDF3nLjd=PXjv6ey6v1OTK=w@mail.gmail.com>
- Spamdiagnosticmetadata: NSPM
- Spamdiagnosticoutput: 1:99
> -----Original Message-----
> From: Alexander Fedotov [mailto:alfedotov@gmail.com]
> Sent: 20 September 2017 16:24
> To: Jon Beniston
> Cc: Tamar Christina; newlib@sourceware.org; nd
> Subject: Re: [AArch64] nosys.specs and printf
>
> >>That's not doing anything other than setting errno to ENOSYS to indicate
> it's not available.
> Yes the same as ARM.
>
> So to sum up. We have several points:
> 1. ARM port doesn't warn user about missing _write(). While this function is
> present in final ELF and does nothing.
> 2. AArch64 warns "warning: _write() is not implemented and will always fail"
> while _write() it is present in ELF.
Both of these are probably correct, while I consider the ARM behaviour to be the incorrect one
It may be so for historical reasons. I'm not really sure.
> 3. libnosys has inconsistent configuration against rdimon. It is impossible to
> switch between rdimon.specs and nosys.specs.
I don't quite get what you mean with this, but I assume you mean that using nosys.specs with AArch64
Results in link errors if you use I/O. Again this is because we don't support nosys for AArch64.
That said, a patch to have libnosys compile (with warnings) for AArch64 would be fine.
Tamar.
>
> Thoughts ?
>
> Alex
>
> On Wed, Sep 20, 2017 at 5:08 PM, Jon Beniston <jon@beniston.com> wrote:
> >> Well at least _write is here:
> >
> > That's not doing anything other than setting errno to ENOSYS to indicate it's
> not available.
> >
> >
>
>
>
> --
> Best regards,
> AF