Re: [PATCH 0/4] Rework isar-apt

[Alexander Smirnov <asmirnov@ilbers.de>]

On 02/01/2018 08:03 PM, Jan Kiszka wrote:
> On 2018-02-01 17:54, Alexander Smirnov wrote:
>>
>>
>> On 02/01/2018 07:25 PM, Jan Kiszka wrote:
>>> On 2018-02-01 17:22, Alexander Smirnov wrote:
>>>> On 02/01/2018 07:14 PM, Jan Kiszka wrote:
>>>>> On 2018-02-01 12:29, Alexander Smirnov wrote:
>>>>>> Hi all,
>>>>>>
>>>>>> this series intended to make buildchroot able to work with isar-apt.
>>>>>> I've tried to add extended comments to each patch.
>>>>>>
>>>>>> Impact: with this series I'm able to build 'example-hello' <->
>>>>>> 'libhello'
>>>>>> without any hacks. So the deps are installed automatically.
>>>>>>
>>>>>> NOTE: I've migrated to bitbake [lockfiles] mechanism, don't know how
>>>>>> robust it
>>>>>> is, but build in the loop started in the evening didn't fail during
>>>>>> the night.
>>>>>>
>>>>>
>>>>> I can happily report: It works as promised also for my use case
>>>>> (jailhouse.bb -> linux-jailhouse.bb, both Isar-built).
>>>>>
>>>>
>>>> Thanks! But I've found an issue with events. :-( Our bitbake didn't
>>>> handle them for multiconfig, the following patch seems to add this:
>>>>
>>>> https://patchwork.openembedded.org/patch/141626/
>>>>
>>>> Events is very good mechanism to clean up Isar build from pending mounts
>>>> without headache with build fails, so I think it would be valuable to
>>>> try latest bitbake. I'll report the results as soon as build finishes.
>>>
>>
>>  From the first like I like it:
>>
>>   - build output looks more user-friendly:
>> 0: mc:qemui386-jessie:example-hello-0.2+7bf716d2-r0 do_build - 222s (pid
>> 15637)
>>
>>   - Yes! Events are now handled by all instances, so no more pending
>> mounts. :-)
> 
> Even on ^C^C (forced termination)? That would be valuable!

It seems so, I've tested several times by interrupting the build with 
double CTRL+C, isar-apt is being unmounted in 1-2 seconds for *all* 
multiconfigs. :-)

So I definitely want this bitbake and add /proc, /dev to clean up hook.

> 
>>
>>> Perfect: Claudius just told me we need to update bitbake anyway to have
>>> proper multiconfig support (i.e. no more hacky copying of files during
>>> isar-init-build-env). Claudius, which version at least?
>>>
>>
>> Hmm, didn't get what is "hacky copying"...
> 
> That we only pick up an use multiconfig/ files from meta-isar, not any
> other layer. See
> https://groups.google.com/forum/#!msg/isar-users/IjQTuBFLPLo/IqRREtAHBgAJ
> 

Ok.

>>
>>> BTW, why are we copying bitbake in? Wouldn't a submodule make more sense?
>>>
>>
>> In my opinion not, the main idea is the same as for Yocto - have no
>> external dependencies. Isar should be self-contained.
> 
> That wouldn't change. You would still have both artifacts in the tree,
> after checkout. And you could tar them both together for release
> packaging. Or what are the concrete downsides?
> 
> I'm not using submodules heavily, but here would be a good use case IMHO.
> 

Have no experience in creation of submodules but sounds good, let me 
check this first in details, thanks for the idea!

Alex