[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#673424: Fwd: Bug#673424: bbswitch packaging



First off, sorry for the slow replies lately...

On Tue, Apr 9, 2013 at 9:03 AM, Aron Xu <aron@debian.org> wrote:
> On Thu, Apr 4, 2013 at 5:22 PM, Vincent Cheng <vincentc1208@gmail.com> wrote:
>> On Wed, Apr 3, 2013 at 5:13 AM, Aron Xu <aron@debian.org> wrote:
>> [snip]
>>> Then could you add it to Debian's git repo?
>>
>> Done. But in the process of building the packages I hit another issue
>> [1], so please hold off (yet again) on uploading primus until it gets
>> fixed.
>>
>
> Do you think it's time to upload bumblebee?

I can now build bumblebee and primus again, but I can't get optirun -b
primus to work properly with the latest packages from git, for some
reason. Have you given those packages a try for yourself yet? I'm
unsure if it's an issue on my end, or something that's expected by
upstream right now. Upstream says that "Rebasing for Bumblebee is less
a good idea, however it's needed to be able to package it using the
current packaging scripts" [1], so I think that they're aware that the
latest code from git is not so stable and we probably shouldn't upload
it as is...

We could also just revert my latest commits in bumblebee + primus git
and just upload packages prior to the removal of primusrun. That'll
mean that we'll have to carry primus-nvidia as a transitional package
in the future, however. It's either this, or we wait until bumblebee
3.2 + primus get released with upstream's current issues sorted out.

Any thoughts?

>>>> As an aside, I made a comment about the current architecture field of
>>>> bbswitch after Ratesh uploaded 0.6, but I suppose you may have missed
>>>> them:
>>>>
>>>> "Also, why did you opt for Architecture: linux-any for a dkms package?
>>>> Everything inside the binary package is installed into an
>>>> arch-independent  location, so I think it should probably be arch:all
>>>> instead, and most dkms packages [1] adhere to being arch:all,
>>>> including dkms itself. But since you've  explicitly moved the package
>>>> from arch:all to arch:linux-any, I'll just leave it be..."
>>>>
>>>
>>> AFAIK even though bbswitch does not contain any architecture specific
>>> file, it does not work on other platforms other than linux-any, e.g.
>>> kfreebsd and hurd. So I moved it to linux-any. (And yes, there is dkms
>>> support for kfreebsd.)
>>
>> However, we end up duplicating the package on all linux archs (there's
>> no difference between the bbswitch package built on i386 vs. amd64, or
>> mips, or sparc, or ppc...). It just feels redundant to me, but on the
>> whole it's just a minor issue. I'm fine with leaving it as-is.
>>
>> How about bumblebee though? That really should be restricted to i386
>> and amd64 only; Nvidia Optimus is AFAIK only supposed to work with
>> Intel+Nvidia hardware combinations, so that pretty much limits it to
>> being used on i386 + amd64.
>
> I guess yes? Don't know other people's opinion.

It's a minor issue either way. /me shrugs

Vincent

[1] https://github.com/Bumblebee-Project/bumblebee-ppa/pull/10


Reply to: