-
Notifications
You must be signed in to change notification settings - Fork 425
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Incorrect library load paths due to rpath patcher run on symlinks #5179
Comments
@h-vetinari, hm.. |
For some reason this started to fail for me after the change from gh-5031 ... |
Resolved in #5181, upcoming |
What happened?
In
conda-build>=3.28.0
, 1dc2470 replaced.os_utils.pyldd.is_codefile
/.os_utils.*.codefile_type
by direct calls to.os_utils.*.codefile_class
which those functions wrapped.For
is_codefile
/codefile_type
theskip_symlinks
parameter defaults toTrue
whereas it isFalse
forcodefile_class
which induced some unintended changes.Notably,
.post.post_process_shared_lib
runs the rpath patcher on symlinks inconda-build>=3.28
, which leads to invalid entries being set for the symlink targets in case the symlinks are located in different directories.Case analysis at conda-forge/openmpi-feedstock#142 (comment) ff.
In that specific case, we have a
lib/<target>
andlib/<subfolder>/<symlink> -> lib/<target>
for whichlib/<target>
gets rpath set to$ORIGIN/..:$ORIGIN/.
instead of$ORIGIN/..:$ORIGIN/.
if the symlink is processed after the actual target.The issue was noticed because we ran into an bug with in
patchelf
which I could pinpoint to being introduced in NixOS/patchelf@42394e8 (i.e., affectingpatchelf>=0.17.0
.(Others have run into issues introduced with that change; see NixOS/patchelf#446 ).
Test and patch incoming.
The text was updated successfully, but these errors were encountered: