Re: [Minios-devel] [UNIKRAFT PATCH] lib/vfscore: sys_ioctl() - Fix build with nolibc

[Sharan Santhanam <sharan.santhanam@xxxxxxxxx>]

On 9/11/19 3:38 PM, Costin Lupu wrote:
> Hi Sharan,
>
> On 9/11/19 3:56 PM, Sharan Santhanam wrote:
> > Hello Costin,
> >
> > The fix seems fine. Please find the question inline.
> >
> > Thanks & Regards
> >
> > Sharan
> >
> > On 9/11/19 1:56 PM, Costin Lupu wrote:
> > > Commit 3dcccd04 introduced handling of FIOCLEX and FIONCLEX requests.
> > > However,
> > > these flags are not defined in nolibc.
> > >
> > > Signed-off-by: Costin Lupu <costin.lupu@xxxxxxxxx>
> > > ---
> > >    lib/nolibc/include/sys/ioctl.h | 0
> > Why do we introduce a empty file? In newlib we introduces a header
> > imported from musl.
> I didn't get that completely, does it build on your side? Do you get any
> errors/warnings?
No, but I guess with a review process you do clarify why you made 
certain choices.
> > Why don't we use the same file here?
> This is an open question. The thing is that newlib itself seems to be a
> poor choice if we do copy so much code from musl. Now getting back to
> nolibc, if we do add more and more code from musl then we can simply get
> rid of it too and use musl instead. In conclusion, I fail to see why we
> should copy code to nolibc instead using musl directly.
But in this case we are introducing the FIONCLEX and FIOCLEX within the 
core Unikraft and it is expected to work with nolibc. Instead of adding 
#ifdef it would be better to make it feature complete.
> Cheers,
> Costin
_______________________________________________
Minios-devel mailing list
Minios-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/minios-devel