Re: wrappers and gcc -isystem

[Jonathan Perkin <jperkin%joyent.com@localhost>]

* On 2016-02-03 at 17:05 GMT, Joerg Sonnenberger wrote:

> On Wed, Feb 03, 2016 at 09:03:01AM +1300, mark wrote:
> > On 2016-02-03 01:33, Joerg Sonnenberger wrote:
> > >*sigh* Why? There are only two reasons for using -isystem over -I:
> > >- include order
> > >- suppression of warnings
> > >
> > >Neither sounds like a good idea for this.
> > 
> > That may be true but it does seem that its increasingly being used out
> > there:
> 
> A lot of people like to hard-code various GCC optimiser flags to
> workaround broken optimisations in compilers from 15 years ago. Doesn't
> mean that it is a good idea. If anything, it makes the idea of
> blacklisting -isystem even stronger.

I'm in full agreement here.  -isystem is the reason why qt5-qtbase is
currently broken on SunOS:

  http://us-east.manta.joyent.com/pkgsrc/public/reports/upstream-trunk32/20160216.2232/qt5-qtbase-5.5.1nb2/build.log

The issue is that the OS stddef.h is being picked up before GCC's due
to "-isystem /usr/include" overriding the default include order, so
::max_align_t is never defined.

It's not just an issue on SunOS, the same problem is causing nodejs to
currently fail on Darwin/gcc when using GCC 4.9 too.

+1 on removing -isystem completely.

-- 
Jonathan Perkin  -  Joyent, Inc.  -  www.joyent.com