Bug 5240 - do_rootfs fails for libcrypto due to stale cache from version change of openssl recipe
Summary: do_rootfs fails for libcrypto due to stale cache from version change of opens...
Status: RESOLVED INVALID
Alias: None
Product: OE-Core
Classification: Build System, Metadata & Runtime
Component: connectivity (show other bugs)
Version: 1.4
Hardware: x86 Multiple
: Low minor
Target Milestone: 1.5.1
Assignee: Cristian Iorga
QA Contact:
URL:
Whiteboard:
Depends on:
Blocks:
 
Reported: 2013-09-19 21:02 UTC by Rob Calhoun
Modified: 2013-09-25 14:32 UTC (History)
4 users (show)

See Also:
OS type for building Yocto: ---
Type of Regression: ---
Verified:
Documentation change: No (bug/feature does not impact docs)


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Rob Calhoun 2013-09-19 21:02:05 UTC
Starting with poky "dylan", build image containing openvpn (custom autotools recipe), which has opensll as a DEPENDS. Upgrade to poky branch 1.4_M5 via git and rebuild. Build of image fails, relevant line being:

Computing transaction...error: Can't install libssl1.0.0-1.0.0j-r15.4@armv7a_vfp_neon: no package provides libcrypto1.0.0 >= 1.0.1e

Fix is "bitbake -c cleanall openssl" followed by re-baking the image.

Core issue appears to be that openssl rev moved backwards in 1.4_M5, with branch 1.4_M5 containing a recipe for openssl_1.0.0j.bb while dylan release had openssl_1.0.1e.bb, resulting in an invalid cache.

http://git.yoctoproject.org/cgit/cgit.cgi/poky/log/meta/recipes-connectivity/openssl?h=1.4_M5

http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/recipes-connectivity/openssl?h=dylan

It looks like the commits since 4fb837687dd68363f25fbfc15207dd05d1369661 had some issues so dropping back to 1.0.0j might have been intentional. Bitbake should invalidate the cache when this happens. Baring making a change there, a comment to the release notes re: the need for a clean build of openssl would be nice.
Comment 1 Richard Purdie 2013-09-25 09:36:26 UTC
I don't think anything went backwards. 1.0.1 (e) > 1.0.0 (j) so the dylan release had a later version than the 1.4M5 milestone release.

Regardless, it shouldn't have failed like that but it wasn't due to a version going backwards...
Comment 2 Rob Calhoun 2013-09-25 14:32:44 UTC
Re: moving backwards you are correct. I had mixed up the Poky 1.4_M5 *branch* with the Poky 1.5_M4 *tag* on branch dylan. (I'm a long-time svn user but new to git.)

I'm going to mark this as invalid, there being no point to ticketing against the 1.4_M5 branch.