Re: [PATCH] targz-img: Ignore tar exit code 1

[Jan Kiszka <jan.kiszka@siemens.com>]

On 05.12.19 20:17, Vijai Kumar K wrote:
> On Thu, Dec 05, 2019 at 07:04:07PM +0100, Jan Kiszka wrote:
>> On 05.12.19 18:44, vijaikumar.kanagarajan@gmail.com wrote:
>>> From: Vijai Kumar K <Vijaikumar_Kanagarajan@mentor.com>
>>>
>>> Sometimes during packaging tar prints a warning "file changed as we read it"
>>> and exits with return code 1. This is observed when we generete both
>>> wic-img and targz-img together. Ignore the error.
>>
>> Ignoring errors is rarely a good idea. In this case, it tells you that
>> you have a bug in your layer that missing a dependency. I know this
>> because we once had the same issue in a product layer. So: NAK
>>
>> Jan
> 
> Ah! Ok. Back to square one I guess. Its currently observed in our downstream setups
> only. I have no good reason to justify it. Reproducing this issue is gonna be difficult.
> Let me see if I can monitor file changes and pinpoint the exact condition.
> But this is going to take a while.
> 
> If it helps, the scenario is IMAGE_TYPE set to wic-img and targz-img is
> inherited so both are executed for a build. Also, on a seperate note

wic may have found a reason to modify the rootfs while integrating it.
So a first shot could be making targz depend on wic completion.

BTW, this sounds similar to
https://gitlab.com/cip-project/cip-core/isar-cip-core/blob/next/classes/wic-targz-img.bbclass
- and that class may have the same issue...

> IMAGE_TYPE's functionality could be extended to provide multiple image types.
> I believe we could specify only one right now. Former is more clean and particularly
> useful if one wants a wic for sd boot and tar for nfsroot. Planning to take this up
> next.

Right, that is of general interest, indeed. Quirin has something planned
around the same use case as well.

Jan

-- 
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux