[oe] cleaning recipes

Frans Meulenbroeks fransmeulenbroeks at gmail.com
Mon Aug 16 11:10:37 UTC 2010


2010/8/16 Koen Kooi <k.kooi at student.utwente.nl>:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 16-08-10 11:30, Frans Meulenbroeks wrote:
>> 2010/8/16 Frans Meulenbroeks <fransmeulenbroeks at gmail.com>:
>>> 2010/8/16 Koen Kooi <k.kooi at student.utwente.nl>:
>>>> -----BEGIN PGP SIGNED MESSAGE-----
>>>> Hash: SHA1
>>>>
>>>> You removed the default openssl version (0.9.8) without bumping PR for
>>>> recipes using that, so now my autobuilder is quite red due to do_rootfs
>>>> failures (among other build errors related to these removals).
>>>
>>> I will reinstate openssl 0.9.8
>>> conf did not show any pinning but i missed the DEFAULT_PREFERENCE = "-1' in 1.0
>>> seems a good plan to upgrade to 0.9.8o though as it fixes some
>>> security vulnerabilities
>>>
>>
>> reverted the commit for removal of 0.9.8
>> I was unaware that a PR bump of recipes using another recipe needed to be done.
>
> For libraries, certainly, as in this case SOVERSION changed, so the
> resulting packagenames changes from
> libssl0.9.8_0.9.8m-r12.0.5_armv7a.ipk to
> libssl1.0.0_1.0.0-r12.2.5_armv7a.ipk. OE removes the first one, but
> other packages still have it in DEPENDS.
>
> And if SOVERSION doesn't change, bumping PR for dependant recipes is
> needed for things that statically link to it (or even dynamically in
> some case).

Tanks for the explanation.
It would be nice if oe would detect those cases and force a rebuild.
>
> The ffmpeg recipes try to list the know offenders that break when
> updating SRCREV, but I guess a bitbake tool (e.g. 'bitbake whatdepends
> foo') would be nice for these kinds of things.

I've asked before about something like that, forgot why it was not
easy/possible.
i also did try a bitbake -g world to get a list of deps, but that also
failed for some reason.

I would expect that if you have the recipes in a parsed format, it
should not be overly difficult.
Anyhow, my pythonese is too limited to implement something like that.

Frans




More information about the Openembedded-devel mailing list