On Fri, Jul 28, 2006 at 10:37:47PM +0100, Ken Moffat wrote:
>
> Certainly, if 'make check'
> is needed as a separate command it will need the BUILD32 stuff, but
> then we might be better reverting to 'make' with BUILD32. I'll take
> a look at the Makefile if I need to after reviewing results from my
> current build.
>
First, the good news - 64-bit bzip2 does run the tests, my script
had acquired a sed from the temp-system chapter.
The tests are _only_ run from the static build (Makefile, not
Makefile-libbz2_so). They are part of a vanilla 'make' because the
targets are
all: libbz2.a bzip2 bzip2recover test
But, for 32-bit multilib we now _only_ make libbz2.a from the
non-shared targets, so we don't test it. I can see the following
options for 32-bit multilib:
1. Continue to only build libbz2.a, delete the phrase "and test it".
2. Make (i.e. libbz2.a, bzip2, bzip2recover, test) so that the tests
run.
3. Make libbz2.a with a note that it can be tested by
make CC="gcc ${BUILD32}" CXX="g++ ${BUILD32}" check
(this will build bzip2 for the test, but presumably not
bzip2recover).
I don't have strong feelings about this - if the tests were ever
likely to show a problem, or if they worked on the shared objects
(hey, I patched that once on lfs-dev), they might be worth doing for
32-bit. If people do want the tests to run, I think option 2 is
more straightforward.
Opinions ?
Ken
--
das eine Mal als Tragödie, das andere Mal als Farce
_______________________________________________
Clfs-support mailing list
[email protected]
http://lists.cross-lfs.org/cgi-bin/mailman/listinfo/clfs-support
