Open nunoguedelha opened 4 years ago
Is this what we are doing when evaluating/selecting the desired versions of osqp?
I am not really sure what CMake's ExternalProject does internally, but at least for the robotology-superbuild on a clean build (so no subsequent tag changes), it seems that the correct version of qdldl is used.
So there would be a problem only if we first clone master then checkout an earlier commit. I get it.
As a side note, git submodule update --recursive
will checkout the proper commit of the submodule. I was reading this tutorial: https://git-scm.com/book/en/v2/Git-Tools-Submodules.
This is is not to point out a real issue, but more a heads up and to verify with you, @GiulioRomualdi , @traversaro , that we are doing the same thing when selecting a given commit of a repo containing git submodules. Let's consider the example of https://github.com/oxfordcontrol/osqp.git (I open the issue here because of
osqp
being an external repo obviously):master
(commit6fb6db9
), the respective commit value for the submodulelin_sys/direct/qdldl/qdldl_sources
is12d56ee
(tagv0.1.4
).if we checkout the tag
v0.6.0
(commit0baddd3
), the respective commit value for the submodulelin_sys/direct/qdldl/qdldl_sources
should be70596af
(commit messageMerge pull request #22 ...
), but in fact, the actual commit checked out in the submodule stays the same:12d56ee
(tagv0.1.4
). This "discrepancy" then appears as a change in the working tree of the parent repoosqp
:no changes added to commit (use "git add" and/or "git commit -a")