Copying in Paul as he was involved in this code so might have some thoughts…

On 9 Oct 2022, at 20:57, Otavio Salvador via lists.openembedded.org 
<[email protected]> wrote:
> 
> Hello all,
> 
> Daiane and I are working on the 3rd edition of our Yocto Project book and now 
> working on the chapter where we'd like to cover the eSDK use better. We are 
> likely missing something obvious, but it is better to ask if we can't figure 
> it out.
> 
> How is 'devtool sdk-install' supposed to work?
> 
> We used the 'devtool sdk-install -s libpng' command, and the build succeeded, 
> but the package didn't install. The error is:
> 
> ERROR: Failed to install libpng - unavailable

Yeah, I just hit that too.

I wonder if the hash equiv broke something…

Interestingly, the eSDK tests use that, so either they’re doing something 
special, or they’re not actually running.  It may be the latter.

It looks like some of the sstate is missing in the eSDK.  For example, 
conf/locked-sigs.inc says:

    
libxml2:do_package_write_rpm:787fe8aabd0010122c8c2070aa0c6e3fac39895963e19600a48226ab3e2b1346

And indeed, this sstate exists:

sstate-cache/78/7f/sstate:libxml2:cortexa57-poky-linux:2.9.14:r0:cortexa57:10:787fe8aabd0010122c8c2070aa0c6e3fac39895963e19600a48226ab3e2b1346_package_write_rpm.tar.zst

However, the populate-sysroot sstate:

    
libxml2:do_populate_sysroot:a37ec72685d847402a8a18b781f453ad30f6aff72892e457e36456163714bb81

Does not exist at all.

So it looks like the eSDK generation is failing. Maybe this is a hash equiv 
issue and the eSDK copy of the sstate is using the wrong hashes?

Ross
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#171572): 
https://lists.openembedded.org/g/openembedded-core/message/171572
Mute This Topic: https://lists.openembedded.org/mt/94223146/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to