On Mittwoch, 16. August 2017 17:33:32 CEST Christian Kandeler wrote:
> On Wed, 16 Aug 2017 16:46:09 +0200
>
> Allan Sandfeld Jensen <k...@carewolf.com> wrote:
> > On Mittwoch, 16. August 2017 11:06:26 CEST Christian Kandeler wrote:
> > > On Tue, 15 Aug 2017 22:09:15 +0200
> > >
> > > Allan Sandfeld Jensen <k...@carewolf.com> wrote:
> > > > On Dienstag, 15. August 2017 16:14:45 CEST Thiago Macieira wrote:
> > > > > On Tuesday, 15 August 2017 03:18:03 PDT Allan Sandfeld Jensen wrote:
> > > > > > You might want to cross compile Qt or most of Qt, but have other
> > > > > > projects
> > > > > > you want to build locally.
> > > > > >
> > > > > > I for one would love such a feature for testing builds on various
> > > > > > build
> > > > > > configuration. So I can cross-build most of Qt, but ensure that we
> > > > > > can
> > > > > > still build the module I am testing, on such an architecture, not
> > > > > > just
> > > > > > for it.
> > > > >
> > > > > If you've done your job right, you can cross compile and that should
> > > > > be
> > > > > superior than building on a slow device.
> > > >
> > > > But it wouldn't verify if I CAN build on a non-x86 architecture, which
> > > > is
> > > > what I am trying to test.
> > >
> > > But what exactly are you testing then? Whether the native compiler
> > > works?
> >
> > In my case we have a tricky translation of qmake settings to GN settings
> > in
> > qtwebengine, so I am testing if everything get translated correctly, so it
> > can actually build. This has broken multiple times. Basically does ARM
> > host builds work or did we accidently make some critical arm part
> > cross-build only.
> I'm intrigued. Can you give an example? I can easily see how things can go
> wrong the other way around ("whoops, we assumed that target equals host"),
> but if you got the target right already, one would think everything's fine.
> Unless it's about the location of host tools...
>
For non-x86 you often have to pass through or trigger the right CFLAGS,
otherwise you don't get the right FP-implementation, architecture revision,
SIMD instructions or ABI. That sort of thing.
Conversely there are often problems with cross-builds if detection goes wrong,
so it starts assuming things detected on host is present on target, which can
make it easier to start by fixing a standard build before fixing the cross-
build, and we have builds breaking regularly due to importing multi-million
lines diffs from chromium.
'Allan
_______________________________________________
Interest mailing list
Interest@qt-project.org
http://lists.qt-project.org/mailman/listinfo/interest
