[Buildroot] Analysis of build results

[Giulio Benetti]

Hi Fabrice,

On 2/22/20 10:11 PM, Fabrice Fontaine wrote:
> Dear all,
> 
> Le sam. 22 févr. 2020 à 20:44, Giulio Benetti
> <giulio.benetti at micronovasrl.com> a écrit :
>>
>> Hi Romain,
>>
>> +Cc me @benettiengineering.com and...
>>
>> Il 22/02/2020 20:10, Romain Naour ha scritto:
>>> Hi Giulio,
>>>
>>> CC: buildroot <buildroot at buildroot.org>
>>>
>>> Le 22/02/2020 à 19:57, Giulio Benetti a écrit :
>>>> Hi Romain,
>>>>
>>>> Il 22/02/2020 19:18, Romain Naour ha scritto:
>>>>> Hi Giulio,
>>>>>
>>>>> Le 21/02/2020 à 00:09, Giulio Benetti a écrit :
>>>>>> Hi Thomas,
>>>>>>
>>>>>> Il 20/02/2020 19:23, Thomas Petazzoni ha scritto:
>>>>>>> On Thu, 20 Feb 2020 19:16:37 +0100
>>>>>>> Giulio Benetti <giulio.benetti at micronovasrl.com> wrote:
>>>>>>>
>>>>>>>> So something like:
>>>>>>>> # while true; do make foo && make foo-dirclean; done
>>>>>>>>
>>>>>>>> but please do you have a 'standard' form where it escapes on error?
>>>>>>>
>>>>>>> while true; do make foo && make foo-dirclean || break ; done
>>>>>>
>>>>>> Unfortunately the problem doesn't show up, I've also accelerated the building
>>>>>> launching a local script in build/libsvgtiny-xxx/ that call make -j17 && make
>>>>>> clean with all the arguments Buildroot provides to accelerate, but nothing. I've
>>>>>> tried on my Workstation, my laptop and 1 of my autobuilders.
>>>>>>
>>>>>> Romain, have you had the chance to try that patch? Can I do something to ease
>>>>>> the thing? Also, which specific distro(Distro name, kernel version) do you have?
>>>>>
>>>>> This a a Fedora 31 disto regularly updated.
>>>>>
>>>>> This is weird since we already disabled parallel build [1]
>>>>
>>>> Ah good finding, with this I've also found that [1] is wrong, since this is a
>>>> generic package, so need to build with $(MAKE1) instead of $(MAKE) without
>>>> LIBSVGTINY_MAKE = $(MAKE1)
>>>> but maybe the patch i've sent you could work.
>>>>
>>>> Can you give a try?
>>>
>>> I tried to reproduce the issue but libsvgtiny still build fine here...
>>>
>>>>
>>>> If it still fails, since this package contains 4 source files, I could send a
>>>> patch that really avoid parallel build. And it must sent anyway a patch to fix
>>>> that error.
>>>
>>> Maybe we can just fix the patch [1] ?
>>> Can you send the patch ?
>>
>> ...yes, I'm going to send it soon.
>>
>>> BTW, who is using libsvgtiny? There is no package using it, no DEVELOPERS entry,
>>> a fragile build system and the version we use is out of date (2011-03-21).
>>> Maybe we can drop it?
>>
>> Yes, it should at least be bumped, I could take care of it, but I don't
>> know if someone uses this package. Anyway I've seen that tags are set
>> per year, so my effort would be very few to bump it and keep it up to date.
> It should be noted that to bump it, you'll also have to bump
> netsurf-buildsystem as well as add libdom package which is used
> instead of libxml2 since
> https://source.netsurf-browser.org/libsvgtiny.git/commit/?id=d03a7471001c4701e1fea976aee162b0de375f52
> libdom depends on libparserutils and libwapcaplet which also need to
> be added to buildroot.
> So that's probably a lot of work for a package that is not used
> currently by any package.

Fortunately you've described all the implications for bumping this 
package, thank you! After this I can tell I won't bump and maintain that 
package.

Well, I've just sent the patch to fix parallel build:
https://patchwork.ozlabs.org/patch/1242561/

Kind regards
-- 
Giulio Benetti
Benetti Engineering sas

>>
>> Thomas, please let me know if it is a good idea or better drop it.
>>
>> --
>> Giulio Benetti
>>
>>> Best regards,
>>> Romain
>>>
>>>>
>>>> Please let me know
>>>> Thank you
>>>>
>>>> Best regards!
>>>
>>
> Best Regards,
> 
> Fabrice
> _______________________________________________
> buildroot mailing list
> buildroot at busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot
>