You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
With py_test (PAR-based) rules I can dev/test iterate on an individual test by running the following:
bazel test //foo:foo_tests
find and update an issue in dependency bar.py
run bazel test //foo:foo_testswithout rebuilding the PAR because the runfiles are symlinked
The iteration cycle (1 and 2 above) is fast because it essentially lays out the zipfile in the runfiles directory and symlinks files that I may want to edit.
With the PEX rules I cannot do this as the only files that are symlinked are the srcs files from the test itself. This means that I have to rebuild the PEX each time I have to edit a dependent file which makes life quite tedious for any serious development.
What should happen is that the runfiles tree is created similarly to what PAR does.
The text was updated successfully, but these errors were encountered:
I agree that would be nice - the challenge lies in keeping all the wheels/eggs/setuptools-stuff dependencies available. We could make a script wrapper that sets PEX_PATH to have pex get dependencies from the previously-built .pex, but I'm not sure how to make it keep a path outside of the .pex in sys.path after pex initializes itself. Any suggestions?
With
py_test
(PAR-based) rules I can dev/test iterate on an individual test by running the following:bazel test //foo:foo_tests
bazel test //foo:foo_tests
without rebuilding the PAR because the runfiles are symlinkedThe iteration cycle (1 and 2 above) is fast because it essentially lays out the zipfile in the runfiles directory and symlinks files that I may want to edit.
With the PEX rules I cannot do this as the only files that are symlinked are the
srcs
files from the test itself. This means that I have to rebuild the PEX each time I have to edit a dependent file which makes life quite tedious for any serious development.What should happen is that the runfiles tree is created similarly to what PAR does.
The text was updated successfully, but these errors were encountered: