bitbake-diffsigs with multiconfig is not helpful

bitbake-diffsigs with multiconfig is not helpful

[chris.laplante]

Hi all,

We are using Yocto Zeus (yes, I know...) and I'm trying to track down some reproducibility issues. My go-to tool is bitbake-diffsigs, but ever since we started using multiconfig, the tool is pretty much useless with results like this:

(where mc-one  and mc-two are multiconfigs)

 NOTE: Starting bitbake server...
runtaskdeps changed:
[-libidn/libidn2_2.2.0.bb:do_package:mc:mc-one libidn/libidn2_2.2.0.bb:do_packagedata:mc:mc-one pseudo/pseudo_git.bb:do_populate_sysroot:virtual:native:mc:mc-one rpm/rpm_4.14.2.1.bb:do_populate_sysroot:virtual:native:mc:mc-one, +libidn/libidn2_2.2.0.bb:do_package:mc:mc-two libidn/libidn2_2.2.0.bb:do_packagedata:mc:mc-two pseudo/pseudo_git.bb:do_populate_sysroot:virtual:native:mc:mc-two rpm/rpm_4.14.2.1.bb:do_populate_sysroot:virtual:native:mc:mc-two]
libidn/libidn2_2.2.0.bb:do_package:mc:mc-one with hash 08ce20399ecf48464970bb3340b5cd5600661a296e679fc260d2bcac29ab2ce8
 changed to
libidn/libidn2_2.2.0.bb:do_package:mc:mc-two with hash a3a359d58c1cbe494935b7689b0a443251ac78666839742a53fd7083969e6e7d
libidn/libidn2_2.2.0.bb:do_packagedata:mc:mc-one with hash 2c5da719673ea0b95c6c8010dad39fae7c6a403d4b72d22786a7caed45606ca3
 changed to
libidn/libidn2_2.2.0.bb:do_packagedata:mc:mc-two with hash 6082ecc1935134fc64f318bf998d65c22b5522aaa6c509026512787495b547f3
rpm/rpm_4.14.2.1.bb:do_populate_sysroot:virtual:native:mc:mc-one with hash 575c6ab8920d8699e291d10ad6211bfae36f4f3b5c682cceb660ef6cfab111e4
 changed to
rpm/rpm_4.14.2.1.bb:do_populate_sysroot:virtual:native:mc:mc-two with hash fed841d6093be9faf415bac2b9dd10f4aafdf6b5bb5075f7da45f2e7028d036d
Dependency on task libidn/libidn2_2.2.0.bb:do_package:mc:mc-two was added with hash a3a359d58c1cbe494935b7689b0a443251ac78666839742a53fd7083969e6e7d
Dependency on task rpm/rpm_4.14.2.1.bb:do_populate_sysroot:virtual:native:mc:mc-two was added with hash fed841d6093be9faf415bac2b9dd10f4aafdf6b5bb5075f7da45f2e7028d036d
Dependency on task libidn/libidn2_2.2.0.bb:do_packagedata:mc:mc-two was added with hash 6082ecc1935134fc64f318bf998d65c22b5522aaa6c509026512787495b547f3
Dependency on task rpm/rpm_4.14.2.1.bb:do_populate_sysroot:virtual:native:mc:mc-one was removed with hash 575c6ab8920d8699e291d10ad6211bfae36f4f3b5c682cceb660ef6cfab111e4
Dependency on task libidn/libidn2_2.2.0.bb:do_packagedata:mc:mc-one was removed with hash 2c5da719673ea0b95c6c8010dad39fae7c6a403d4b72d22786a7caed45606ca3
Dependency on task libidn/libidn2_2.2.0.bb:do_package:mc:mc-one was removed with hash 08ce20399ecf48464970bb3340b5cd5600661a296e679fc260d2bcac29ab2ce8


The next thing to try is diffsigs on the do_package or the do_populate_sysroot siginfos. But so far in my testing, the rabbit hole never ends (granted I've only gone down 6 or 7 levels before giving up). I'm going to cobble together an automated tool to follow the trail for me, but I thought I'd post this first to see if anyone has any ideas.

I have a few questions:
1. I would that the name of the multiconfig (e.g. mc-one vs mc-two) doesn't factor into the hash calculation. Would I be correct?
2. Is anyone aware of the situation improving in newer versions of Yocto?
3. We are also using hash equivalence. Could this have any effect on what I'm seeing?


Thanks,
Chris