Re: m68k buildd presents significant problem for non-us and testing
On Thu, Feb 22, 2001 at 09:52:45AM +0100, Paul Slootman wrote:
> On Wed 21 Feb 2001, Richard Atterer wrote:
> > BTW, has anybody considered setting up a cross-compiling buildd? I
> > find that cross-compiling arm packages on my x86 works very well and
> > is as easy as native compilation (after having dealt with a few basic
> > problems).
>
> It's not that simple, because often, during the build of a package,
> an executable that was compiled just before is used in the build
> process, and such situations won't work.
The build process can detect whether you're cross-compiling and build
things with one or the other compiler as needed. Of course, those
checks for $DEB_BUILD_ARCH != $DEB_HOST_ARCH are not present now, but
they could be added over time.
> However, for a large amount of packages I guess it could work, and
> the remaining packages could be left for the "real" m68k systems to
> build.
Indeed, that's exactly what I was thinking. There is a much more
common problem than that of packages which build binaries for use
during compilation:
Some configure scripts test for features which dpkg-cross does not
provide an answer for, or even worse they use AC_TRY_RUN. But even
this problem can be overcome by executing the configure script (and
*only* the configure script) natively, copying the generated
config.status back to the build machine and execute it there to
generate the Makefiles. This copying etc. can of course also be
automated.
Bye,
Richard
--
__ _
|_) /| Richard Atterer | CS student at the Technische | GPG key:
| \/¯| http://atterer.net | Universität München, Germany | 888354F7
¯ ´` ¯
Reply to: