[OE-core] [PATCH 1/1] opkg 0.1.8: respect to the arch when choose the alternatives

[Koen Kooi]

Op 26 mei 2012, om 08:28 heeft Martin Jansa het volgende geschreven:

> On Sat, May 26, 2012 at 10:47:31AM +0800, Robert Yang wrote:
>> 
>> 
>> On 05/25/2012 07:30 PM, Martin Jansa wrote:
>>> On Fri, May 25, 2012 at 01:19:55PM +0200, Koen Kooi wrote:
>>>> 
>>>> Op 25 mei 2012, om 12:02 heeft Robert Yang het volgende geschreven:
>>>> 
>>>>> There is a bug if we:
>>>>> 1) bitbake core-image-sato-sdk MACHINE=qemux86
>>>>> 2) bitbake core-image-sato with MACHINE=crownbay
>>>>> 
>>>>> Then several pkgs in deploy/ipk/i586 would be installed to crownbay's
>>>>> image even if there is one in deploy/ipk/core2 and we have set the
>>>>> core2's priority higher than i586, when the version in deploy/ipk/i586 is
>>>>> higher. This doesn't work for us, for example, what the crownbay need is
>>>>> xserver-xorg-1.9.3, but it installs xserver-xorg-1.11.2.
>>>>> 
>>>>> This is caused by opkg's selecting mechanism, if there are more than one
>>>>> candidates which have the same pkg name in the candidate list, for
>>>>> example, the same pkg with different versions, then it will use the last
>>>>> one which is the highest version in the list, this doesn't work for us,
>>>>> it should respect to the arch priorities in such a case.
>>>> 
>>>> This is a serious break with the current opkg behaviour and I don't think it's an improvement. Needing different versions for non machine specific packages indicates a more serious bug elsewhere.
>>> 
>>> It's not the same use-case as those 2 above, but what I don't like on
>> 
>> Hi Martin,
>> 
>> They are the same cases:-), I think that this patch has also fixed your problem,
> 
> No, at least not completely the same.
> 
> I would prefer to upgrade foo-1.0-r1_armv4t temporary until
> foo-1.0-r1_armv7a gets available in feed and that won't happen with your
> patch AFAIK.
> 
> with your patch:
> If you have bar-1.0 which has to be MACHINE_ARCH and in 2.0 bar
> developers find way to detect and use all machine capabilities in
> runtime, recipe maintainer will switch to TUNE_ARCH, then 
> foo-1.0_nokia900.ipk won't be ever upgraded to foo-2.0_armv7a.ipk 
> and that's bad.

And what's worse, the cited usecase is for (slightly paraphrasing):

xserver-xorg 1.11.2 i586
xserver-xorg 1.9.3 i686

Which indicates there is a different, more serious problem at hand. It seems that someone is trying to encode machine specific tweaks to non-machine specific packages. I'm more interested in solving that problem than in changing opkg/rpm behaviour.

There are a number of things that are just not possible to do when supporting multimachine builds and/or multimachine feeds. For example:

machine dogbeachmountain (i686) needs xf86-video-evilpowervr-closedbinary that only works with Xorg 1.9, but machine cherryblossomhighway (i586) can use xf86-video-intel with any xserver-xorg.
If you want both of these machines to work in multimachine builds and/or feeds, you need to lock down xserver-xorg to 1.9 for i*86. If you don't want to lock it down and imgtec won't give you a better binary drop, too bad, stop doing multimachine.
I'm not saying the above situation is what Robert is trying to solve, but it's a situation meta-ti is currently facing with the new binary drop for SGX support. When you have dealt with SGX blobs everything starts to look like an SGX problem :) 

regards,

Koen