All, I'm doing some basic sstate testing as follows (note: strace is just an example package used): $ cd /opt/yocto/cache-build/p1020rdb $ source oe-init-env-build $ bitbake strace $ cd /opt/yocto/cache-test/p1020rdb $ source oe-init-env-build $ cp /opt/yocto/cache-build/p1020rdb/build/sstate-cache/* sstate-cache/ $ bitbake strace I'm left with two instances of some sstate-cache: $ bitbake-diffsig sstate-cache/sstate-strace-ppce5500-fsl-linux-4.5.20-r2-ppce5500-2-*_deploy-rpm.tgz.siginfo Dependency on task /opt/yocto/cache-test/meta/recipes-devtools/perl/perl_5.12.3.bb.do_package was added Dependency on task virtual:native:/opt/yocto/cache-test/meta/recipes-devtools/pseudo/pseudo_1.1.1.bb.do_populate_sysroot was added Dependency on task /opt/yocto/cache-test/meta/recipes-devtools/strace/strace_4.5.20.bb.do_package was added Dependency on task virtual:native:/opt/yocto/cache-test/meta/recipes-devtools/rpm/rpm_5.4.0.bb.do_populate_sysroot was added Dependency on task virtual:native:/opt/yocto/cache-build/p1020rdb/meta/recipes-devtools/pseudo/pseudo_1.1.1.bb.do_populate_sysroot was removed Dependency on task /opt/yocto/cache-build/p1020rdb/meta/recipes-devtools/strace/strace_4.5.20.bb.do_package was removed Dependency on task /opt/yocto/cache-build/p1020rdb/meta/recipes-devtools/perl/perl_5.12.3.bb.do_package was removed Dependency on task virtual:native:/opt/yocto/cache-build/p1020rdb/meta/recipes-devtools/rpm/rpm_5.4.0.bb.do_populate_sysroot was removed It seems like just the task paths are changing which should not really be an issue? Although I admit it could for some packages. This seems to be happening for a lot of packages I'm building as well and not just strace.
I've posted a couple of patches on bitbake-devel regarding this. The bottom line is that the path changes were cosmetic, they did show up in diffsigs but the core hashes didn't use them. I'd therefore like to understand whats different after we filter that stuff out of the equation...
(In reply to comment #1) > I've posted a couple of patches on bitbake-devel regarding this. The bottom > line is that the path changes were cosmetic, they did show up in diffsigs but > the core hashes didn't use them. I'd therefore like to understand whats > different after we filter that stuff out of the equation... I think we can mark this NOTABUG - since the hashes where wrong in my case and I've fixed some of these through other means (by correctly adding vardepsexclude to vars). Also, your bitbake patches clear this up for future encounters. -M
The diffisgs issue is resolved in http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=cb66ebaa8bd950c9a58558ee4d74e8f4f6777403 so I'm going to class this bug as fixed.