| Message ID | 20170427053609.GA20301@waldemar-brodkorb.de |
|---|---|
| State | Accepted |
| Commit | 43155c5b4c458c4526be47834c5e99bd11245d56 |
| Headers | show |
>>>>> "Waldemar" == Waldemar Brodkorb <wbx@openadk.org> writes: > The output detection recognized wrong target output, because > sparcv9 optimization flags used for sparcv8 build. > Fixes: > http://autobuild.buildroot.net/results/1b3158b03f7eaf5afb5a4dab9526091888f6c9b8 > Signed-off-by: Waldemar Brodkorb <wbx@openadk.org> Committed, thanks. Don't forget to submit it upstream as well.
Hello, On Thu, 27 Apr 2017 07:36:09 +0200, Waldemar Brodkorb wrote: > +- sparc-* | sparc64-*) > ++ sparc64-*) > + AC_DEFINE([ARCH_SPARC],,[sparc architecture]) > + TRY_CFLAGS="$OPT_CFLAGS -mcpu=ultrasparc -mvis" > + AC_TRY_CFLAGS([$TRY_CFLAGS $CFLAGS],[OPT_CFLAGS="$TRY_CFLAGS"]);; I am not sure this is the completely correct solution, and Arnout proposed another solution, see: https://patchwork.ozlabs.org/patch/749138/ https://patchwork.ozlabs.org/patch/749137/ The second patch has the explanation why Waldemar patch is not really correct. Best regards, Thomas
>>>>> "\Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes: > Hello, > On Thu, 27 Apr 2017 07:36:09 +0200, Waldemar Brodkorb wrote: >> +- sparc-* | sparc64-*) >> ++ sparc64-*) >> + AC_DEFINE([ARCH_SPARC],,[sparc architecture]) >> + TRY_CFLAGS="$OPT_CFLAGS -mcpu=ultrasparc -mvis" >> + AC_TRY_CFLAGS([$TRY_CFLAGS $CFLAGS],[OPT_CFLAGS="$TRY_CFLAGS"]);; > I am not sure this is the completely correct solution, and Arnout > proposed another solution, see: > https://patchwork.ozlabs.org/patch/749138/ > https://patchwork.ozlabs.org/patch/749137/ > The second patch has the explanation why Waldemar patch is not really > correct. Ok. Can I take this as an acked-by for the check-bin-arch / pkg-generic changes then? ;)
Hello, On Thu, 27 Apr 2017 11:15:34 +0200, Peter Korsgaard wrote: > > I am not sure this is the completely correct solution, and Arnout > > proposed another solution, see: > > > https://patchwork.ozlabs.org/patch/749138/ > > https://patchwork.ozlabs.org/patch/749137/ > > > The second patch has the explanation why Waldemar patch is not really > > correct. > > Ok. Can I take this as an acked-by for the check-bin-arch / pkg-generic > changes then? ;) I haven't reviewed the implementation details, but on the general principles, yes, I'm OK. It's a bit annoying to extend the package infrastructure for such a specific case, but I don't really see a better solution here, except hacking libmpeg2 to remove the runtime detection logic. On the other hand, do we care enough about libmpeg2 on SPARC to extend the package infrastructure just for this purpose? In the end, even if Waldemar's solution is not "correct", it avoids the need to extend the package infrastructure just for a purpose of a single package that will most likely never ever be used on SPARC. Arnout, what do you think? Best regards, Thomas
On 27-04-17 11:22, Thomas Petazzoni wrote: > Hello, > > On Thu, 27 Apr 2017 11:15:34 +0200, Peter Korsgaard wrote: > >> > I am not sure this is the completely correct solution, and Arnout >> > proposed another solution, see: >> >> > https://patchwork.ozlabs.org/patch/749138/ >> > https://patchwork.ozlabs.org/patch/749137/ >> >> > The second patch has the explanation why Waldemar patch is not really >> > correct. >> >> Ok. Can I take this as an acked-by for the check-bin-arch / pkg-generic >> changes then? ;) > > I haven't reviewed the implementation details, but on the general > principles, yes, I'm OK. It's a bit annoying to extend the package > infrastructure for such a specific case, but I don't really see a > better solution here, except hacking libmpeg2 to remove the runtime > detection logic. > > On the other hand, do we care enough about libmpeg2 on SPARC to extend > the package infrastructure just for this purpose? In the end, even if > Waldemar's solution is not "correct", it avoids the need to extend the > package infrastructure just for a purpose of a single package that will > most likely never ever be used on SPARC. > > Arnout, what do you think? When I spun that patch, I expected there would be more use cases than just libmpeg2. But the autobuilders don't point to anything obvious, so perhaps not. I've marked as Rejected in patchwork. We can pick it up again if something else turns up. Regards, Arnout
>>>>> "Arnout" == Arnout Vandecappelle <arnout@mind.be> writes: >> Arnout, what do you think? > When I spun that patch, I expected there would be more use cases than just > libmpeg2. But the autobuilders don't point to anything obvious, so perhaps not. > I've marked as Rejected in patchwork. We can pick it up again if something else > turns up. Ok, sounds sensible - Thanks!
Hi Peter, Peter Korsgaard wrote, > >>>>> "Waldemar" == Waldemar Brodkorb <wbx@openadk.org> writes: > > > The output detection recognized wrong target output, because > > sparcv9 optimization flags used for sparcv8 build. > > > Fixes: > > http://autobuild.buildroot.net/results/1b3158b03f7eaf5afb5a4dab9526091888f6c9b8 > > > Signed-off-by: Waldemar Brodkorb <wbx@openadk.org> > > Committed, thanks. Don't forget to submit it upstream as well. Upstream looks very inactive to me, I don't think I will submit it there. best Waldemar
diff --git a/package/libmpeg2/0004-fix-sparc.patch b/package/libmpeg2/0004-fix-sparc.patch new file mode 100644 index 0000000..d876b66 --- /dev/null +++ b/package/libmpeg2/0004-fix-sparc.patch @@ -0,0 +1,16 @@ +Do not use sparcv9 optimization flags for sparcv8 builds + +Signed-off-by: Waldemar Brodkorb <wbx@openadk.org> + +diff -Nur libmpeg2-0.5.1.orig/configure.ac libmpeg2-0.5.1/configure.ac +--- libmpeg2-0.5.1.orig/configure.ac 2008-07-18 16:30:17.000000000 +0200 ++++ libmpeg2-0.5.1/configure.ac 2017-04-26 21:09:15.780838339 +0200 +@@ -95,7 +95,7 @@ + break + fi + done;; +- sparc-* | sparc64-*) ++ sparc64-*) + AC_DEFINE([ARCH_SPARC],,[sparc architecture]) + TRY_CFLAGS="$OPT_CFLAGS -mcpu=ultrasparc -mvis" + AC_TRY_CFLAGS([$TRY_CFLAGS $CFLAGS],[OPT_CFLAGS="$TRY_CFLAGS"]);;
The output detection recognized wrong target output, because sparcv9 optimization flags used for sparcv8 build. Fixes: http://autobuild.buildroot.net/results/1b3158b03f7eaf5afb5a4dab9526091888f6c9b8 Signed-off-by: Waldemar Brodkorb <wbx@openadk.org> --- package/libmpeg2/0004-fix-sparc.patch | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) create mode 100644 package/libmpeg2/0004-fix-sparc.patch