vtk update missing migrator

[jakirkham]

A few weeks back vtk was updated to 9.3.0: conda-forge/vtk-feedstock#328

However the bot did not open a migrator PR

So @minrk needed to open one manually: conda-forge/conda-forge-pinning-feedstock#6046

Do we know why the bot missed adding this migrator for vtk version 9.3.0?

[beckermr]

We lost an opportunity to debug when the pr was opened manually. Maybe I can cook up a test, but it will be a bit.

[jakirkham]

Ah ok. Sorry still catching up a bit

Maybe we can keep an eye on this for the next vtk update

[jakirkham]

Looks like another debugging window is opening with vtk version 9.3.1

xref: conda-forge/vtk-feedstock#330

[minrk]

I'll make sure to not do the manual migrator this time.

[minrk]

I think the rerender error in the 9.3.1 PR probably points to the culprit, a recursion error in hash_dependencies:

File "/opt/conda/lib/python3.11/site-packages/conda_build/metadata.py", line 1787, in hash_dependencies
    hashing_dependencies = self.get_hash_contents()
                           ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/lib/python3.11/site-packages/conda_build/metadata.py", line 1734, in get_hash_contents
    dependencies = set(self.get_used_vars())
                       ^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/lib/python3.11/site-packages/conda_build/metadata.py", line 2869, in get_used_vars
    script_reqs = self._get_used_vars_output_script()
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/lib/python3.11/site-packages/conda_build/metadata.py", line 2988, in _get_used_vars_output_script
    self.get_rendered_output(self.name(), permit_undefined_jinja=True) or {}
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/lib/python3.11/site-packages/conda_build/metadata.py", line 2821, in get_rendered_output
    for output_ in self.get_rendered_outputs_section(
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/lib/python3.11/site-packages/conda_build/metadata.py", line 2803, in get_rendered_outputs_section
    self._get_contents(
  File "/opt/conda/lib/python3.11/site-packages/conda_build/metadata.py", line 2054, in _get_contents
    env.globals.update(get_dict(m=self, skip_build_id=skip_build_id))
                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/lib/python3.11/site-packages/conda_build/environ.py", line 370, in get_dict
    d.update(meta_vars(m, skip_build_id=skip_build_id))
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/lib/python3.11/site-packages/conda_build/environ.py", line 597, in meta_vars
    d["PKG_HASH"] = meta.hash_dependencies()
                    ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/lib/python3.11/site-packages/conda_build/metadata.py", line 1787, in hash_dependencies
    hashing_dependencies = self.get_hash_contents()

[jakirkham]

Thanks Min! 🙏

Interesting, thought we had fixed a lot of the re-rendering issues recently ( conda-forge/vtk-feedstock#327 ). Though I guess new issues cropped up?

[minrk]

FWIW, I've been testing local builds of conda-smithy and conda-build for conda-forge/conda-smithy#1968, and the recursion error does not happen with latest conda-smithy (conda-forge/conda-smithy@dd6091b), but it does with 3.36.2. I don't really understand which change would cause that, but pushing a conda-smithy release ought to fix it.