Re: [PATCH] Assume that _DIRENT_HAVE_D_TYPE is always defined.

[Florian Weimer <fweimer at redhat dot com>]

On 10/30/2017 04:10 PM, Carlos O'Donell wrote:
> On 10/30/2017 07:46 AM, Florian Weimer wrote:
> > > Our shared source file list says this comes from 'gnulib':
> > > https://sourceware.org/glibc/wiki/SharedSourceFiles
> > >
> > > Are we able to harmonize this with gnulib?
> >
> > Not sure if this is a good idea.  We use the generic code as a
> > building block for the Linux implementation, and I doubt the gnulib
> > implementation will support such abuse.
>
> We have two high-level options:
>
> (a) Approach gnulib with a patch that removes the conditional on
>      _DIRENT_HAVE_D_TYPE, and make the case.

Solaris doesn't have d_type, I think, so that's not really an option for 
them.

> (b) Stop sharing this source file with gnulib, update the SharedSourceFiles
>      not to list the file. Provide a comment in the file stating all of the
>      historical details and why we are not sharing anymore.

If we do that, we should remove the other untested clutter as well, and 
the d_type cleanup would just be a tiny part of that.

I expect that we can't import gnulib getcwd because the code calls 
system getcwd under certain circumstances, and that's obviously a bad 
idea when we use it to implement system getcwd.  We could compile it 
with the appropriate preprocessor flags so that this does not happen, 
but our own goals and gnulib's clearly have diverged by this point.

Thanks,
Florian