Re: [Mingw-cross-env-list] Re: URL problems

[Tony Theodore]

On 28 April 2010 19:22, Volker Grabsch <address@hidden> wrote:
>Persistent URLs are only for whiners who can't accept that 404 is part of life.

QOTW

> Tony, do you think we can switch back to MinGW and use GCC 4.5.0?
>
> Or did you get MinGW-w64 to work?
>
> Or do you recommend to adjust our URLs and stick with GCC 4.4.0
> for this release?

GCC 4.5 seems to basically be a drop in replacement at this stage with
the MinGW runtime/api. Looking at the TDM download page, they have a
previous version of w32api than the one we use, a cvs patched pthreads
(which is easy to prepare), and cloog/ppl (which we don't use at the
moment). There's nothing there really to justify them having a 4.5
release. The 4.4.3 probably wouldn't be worth it at this stage -
unless some project *really* doesn't build with 4.5.

In my previous testing with mingw-w64, it was hard to tell what the
failures were about - GCC 4.5, the api, or the package itself. The
brief testing I did the other day with MinGW shows that all of
mingw-cross-env builds, apart from openscengraph. This leads me to
believe there's nothing (or not much) that is GCC 4.5 or project
specific.

The tests we have pass or fail the same under either 4.4.0 or 4.5. The
gtk and qt tests pass, the others just flash up command windows and I
haven't taken the time to look at them - but the results aren't
related to the gcc version. Additionally, most of the qt demos run
perfectly, one fails due to the sqlite driver, the other pops up a
message along the lines of "This demo is GPU/CPU intensive and may not
work on you system". Again, these are the same across gcc versions.

So I'd say mingw-w64 still has a way to go with their runtime/api, but
switching to GCC 4.5 means that we can compare them very easily. We
can get the i686 target working with MinGW as a reference, then there
should only be 64 bit issues left over (since they share the same
codebase).

All in all, I'm fairly optimistic about switching over to GCC 4.5 with
MinGW. We could revert glib to the 2.4 release (since it's a time sink
- though interesting and worthwhile), and spend the time on the
openscengraph problem. The glib seems to be cross-compile/platform
dependent, whereas osg is more compiler/configure related and should
only need to fixed once.

Cheers,

Tony