On Wed, 2017-02-01 at 13:44 +0200, Jussi Kukkonen wrote: > On 1 February 2017 at 13:23, Richard Purdie <richard.purdie@linuxfoun > dation.org> wrote: > > On Wed, 2017-02-01 at 12:03 +0200, Jussi Kukkonen wrote: > > > Recipe-specific sysroots broke make_relative_symlink(), which > > > turns absolute symlinks in sysroots into relative ones. Use the > > > difference between the (in-sysroot) paths to construct the > > relative > > > symlink. > > > > > > This fixes links in openssl-native, fontconfig-native and bzip2- > > > native. > > > > > > Signed-off-by: Jussi Kukkonen <[email protected]> > > > --- > > > > > > sstate is not an area I'm familiar with, please take a good look. > > > > > > As far as I could see outputpath (based on state[2]) was never > > really > > > needed so I did not use it in the new version. > > > > I don't think we can hardcode workdir into here as for tasks like > > do_deploy, this makes no sense. I think we removed most of the > > absolute > > links from the deploy tasks so we currently don't need this, at > > least > > in the common case but the sstate code is meant to be generic. > > > > I am wondering if we need to pass in anything at all, can't we just > > call relpath on the original path and turn it into a relative one > > directly without referencing it back to TMPDIR/WORKDIR? > > > The actual file path during do_populate_sysroot is something like > /mnt/extra-ssd/tmp/work/x86_64-linux/openssl-native/1.0.2j- > r0/sysroot-destdir/mnt/extra-ssd/tmp/work/x86_64-linux/openssl- > native/1.0.2j-r0/recipe-sysroot-native/usr/lib/ssl/certs > > and the link before make_relative_symlink() points to > /mnt/extra-ssd/tmp/work/x86_64-linux/openssl-native/1.0.2j- > r0/recipe-sysroot-native/etc/ssl/certs > > Assuming those are correct, I don't see how to do this without > referencing WORKDIR or TMPDIR?
Good point, I knew I was missing something. How about passing in walkroot/state[1] into the function, then you can subtract that from the actual file, then run relpath between the (srcpath - walkroot) and the link? Cheers, Richard -- _______________________________________________ Openembedded-core mailing list [email protected] http://lists.openembedded.org/mailman/listinfo/openembedded-core
