Re: PPC64 port status
On Wed, Mar 30, 2016 at 04:41:40PM +0200, Mathieu Malaterre wrote:
> On Wed, Mar 30, 2016 at 4:30 PM, Lennart Sorensen
> <lsorense@csclub.uwaterloo.ca> wrote:
> > On Wed, Mar 30, 2016 at 04:26:59PM +0200, Mathieu Malaterre wrote:
> >> Let me rephrase that in: "Wow somebody (user)tagged specifically with
> >> ppc64 the nouveau bugs" ... :)
> >>
> >> This was not clear. But the offer to ppc64 machine(s) was made by the
> >> ppc64el people:
> >>
> >> https://lists.debian.org/debian-powerpc/2015/06/msg00015.html
> >>
> >> As you mentioned this is non-trivial to get access to a ppc64 machine otherwise.
> >
> > Certainly. And getting access to one of the ones that causes trouble
> > by not having VMX seems even harder in fact.
> >
> > I wonder if there is any easy way to scan binaries for VMX instructions
> > to see which ones actually need rebuilding.
>
> Interesting. Debian BTS is full of reports from people using 586 cpu
> where SSE instructions makes the process crash[*]. So if anything can
> be done to detect that at PPC64 level it should be propagated to 586
> people...
>
> 2cts
> [*] https://bugs.debian.org/764291
Of course such a tool will have false positives from any code that
has runtime detection and multiple code paths for such cpu differences
(I suspect libc will have that for example). But at least it could tell
you which packages are not a problem.
--
Len Sorensen
Reply to: