Add workflow using conda to install dependencies #313
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
diegoferigo
force-pushed
the
feature/conda
branch
from
4e09529 to
8d44bb1
Compare
In general FYI I am all in favor in progressively moving dependencies with a clear versioning API/ABI and release policy from the robotology channel to proper conda-forge dependencies. iDynTree is definitely one of the first candidates, the missing part is just to move the idyntree-yarp-tools to https://github.com/robotology/idyntree-yarp-tools . |
My comment was not clear enough: substitute |
Actually I did not even noticed the conda-forge/robotology typo. I just mentioned that for mutual alignment. |
diegoferigo
force-pushed
the
feature/conda
branch
from
8d44bb1 to
013bb35
Compare
6 tasks
diegoferigo
force-pushed
the
feature/conda
branch
3 times, most recently
from
dc8de7f to
5a3f611
Compare
diegoferigo
force-pushed
the
feature/conda
branch
from
a8f0141 to
76424c7
Compare
|
Rebased and improved the workflow. Still blocked by conda-forge/libignition-gazebo-feedstock#4. |
|
I re-triggered CI just to check and now after conda-forge/libignition-common-feedstock#36, the error of the wrong plugin directory seems solved. The project is building (also before it was) and the tests run and fail. This time, the error is the following:
Inspecting the conda environment, no ODE package is installed automatically (this was what I was mentioning today about the runtime dependencies of ignition libraries packaged in conda-forge @traversaro). mamba list |
|
Now few tests are failing, but the majority are passing. Definitely a step forward. I'm not sure what's missing now, I guess I need to debug it locally, from CI it's not clear to me. Failing tests:
|
diegoferigo
force-pushed
the
feature/conda
branch
from
1fc7edb to
3d75dcc
Compare
|
Rebasing on latest Edit: no changes. |
diegoferigo
force-pushed
the
feature/conda
branch
from
3d75dcc to
256bbf7
Compare
diegoferigo
force-pushed
the
feature/conda
branch
from
256bbf7 to
ef28b84
Compare
diegoferigo
force-pushed
the
feature/conda
branch
5 times, most recently
from
1a9ddcc to
d7f7696
Compare
|
This PR seems now working on linux, but there's a test failing probably due to the dart version that is shipped in conda-forge (I disabled it for the moment). MacOS segfaults during pytest.Windows errors during CMake configuring |
|
Updated status:
On windows, it seems that it links the dll instead of the libs. cc @traversaro any clue? I can see that the library is found as part of the macos: segfaults when scenario is importedwindows: fails to compile due to link error with zlib.dll |
|
I managed to get a more descriptive error message on macos: It seems to pick the |
|
With the workaround provided by @traversaro in conda-forge/libignition-gazebo-feedstock#30 (comment), the windows build compiles until the SWIG bindings, that now fail with: The workflow is compiling in Debug and this could be the cause of this |
|
Compiling in Release seems fixing the SWIG bindings: but now the vendored Physics system is failing with: Trying to use |
|
Now also windows compiles the project successfully. Let's see if we manage to reach the testing step. Similarly to macOS, I suspect that importing the bindings does not work out of the box also on windows. |
diegoferigo
force-pushed
the
feature/conda
branch
2 times, most recently
from
080c6a7 to
343a61d
Compare
Install explicitly libode Disable failing tests due to conda-forge's dart [remove]
[remove] temporarily disable check for ign command
diegoferigo
force-pushed
the
feature/conda
branch
3 times, most recently
from
f90906e to
6ae5e86
Compare
diegoferigo
force-pushed
the
feature/conda
branch
from
6ae5e86 to
6795719
Compare
|
Updated status:
macos: segfaults when scenario is importedwindows: tests fail |
|
@traversaro reading better the windows error:
it seems that something related to the relocation here failed. The file |
Yes, the problem is probably not here, I guess for now you can skip Windows and macOS after just opening an issue in the ign-gazebo feedstock repo on those problems. |
1 task
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
#279 provides useful information about why relying on conda could be beneficial over the current PyPI-based distribution.
Much work has been done by the conda-forge community (including @traversaro) to provide the Ignition stack:
This PR explores the integration of conda and its channels (particularly, conda-forge and robotology) to install the build dependencies in CI.
CMAKE_PREFIX_PATHfor iDynTree does not work in this setup, I had to rely oniDynTree_DIR(note that iDynTree exported targets are inside the python package)This PR is currently blocked by conda-forge/libignition-gazebo-feedstock#4.