Re: [Qemu-devel] [PATCH 00/28] Memory API for 1.6: fix I/O port endianness mess

[Anthony Liguori]

Benjamin Herrenschmidt <address@hidden> writes:

> On Thu, 2013-07-25 at 10:40 +0200, Paolo Bonzini wrote:
>> In any case, let me reinforce point 6 above.  The patches were not
>> reverted because nobody posted the patches to do so.  Until someone
>> does
>> so, it is basically Anthony's call.  If he trusts people to fix the
>> whole thing, he has no reason to revert anything.  My track record of
>> fixing mess is probably better than my track record of not causing it,
>> so he didn't revert it.
>
> That's fine, I know you can fix stuff :-) I'm just really annoyed that
> upstream qemu remained broken for so long (and still is) while the whole
> thing derailed into a mostly pointless discussion on endianness and
> nobody (including Alexey) hollered loud enough that the breakage was
> fairly extensive

It you want to prevent breakages, add unit tests.  No patch will ever be
merged into the tree if 'make check' fails.  It's the ultimately way to
prevent things like this from happening.

I'd rather the tree is broken for a week and we get the proper fix
merged than go through a big production of reverting things.

Regards,

Anthony Liguori

> (BTW. The EHCI problem doesn't seem limited to SLOF, from what I
> remember of what Alexey showed me today, the Linux EHCI driver also
> segfaults qemu).
>
> Cheers,
> Ben.