This involves moving the report files, but it allows me to delete the
indirect variable and indirect array code in update-codgen. As proud as
I was of figuring that out, I am also ashamed of myself for doing it.
This is my atonement.
Use the "subprojects" aspect of update-codegen to generat openapi for
the subprojects. Next we can simplify and remove the generic support.
apiextensions-apiserver seems like it was ALWAYS broken:
k8s.io/apiextensions/ doesn't exist, but k8s.io/apiextensions-apiserver
does.
Fixing that causes different openapi results, obviously.
k8s_tag_files_matching looks for a slash after its argument, so the current value doesnt match anything
also update codegen
this is required for apiextensions-apiserver tests. After fixing apiextensions server tests to use type-aware SSA (instead of erroneously using untyped SSA) there were errors since none of the apiextensions types were actually used in the openapi given to tests.
This moves the hack/ directory and scripts to the examples dir, which is
a distinct module. This avoids some Go unpleasantness around module
boundaries and just makes more sense.
LC_ALL is always wanted and GREP_OPTIONS is never wanted. The `grep
--color=never` dates back to 2016, an issue with OLD grep on Macs, which
was hard to deal with when this was all Makefile magic. Now that it's a
script, we can do it simpler.
Swagger "docs" are actually Go code, which is used by other codegen
tools, so if you really want to regen EVERYTHING, this is part of it and
sequence matters.
Calling update-codegen.sh with no arguments runs all the functions in
definition order. Client-generation depends on protobuf, so protobuf
codegen needs to be near the beginning.
Also add some debug output for protobuf generation.
Also hide some old, verbose debug output.
The `find` tool has hard to comprehend syntax and does not consider
things excluded by .gitignore. I keep tripping over this in my own
repos, where I have __stuff which gets found.
This converts update-codegen to use `git ls-files` in a seemingly
equivalent way (`-cmo --exclude-standard`). I verified it finds the
same set of files as before.
This also drops some obsolete filtering.
Also hide grep errors for not-found files, which can happen if a file is
removed but git ls-files still knows it.
Re-running update-codegen shows no diffs.
This will make subsequent changes easier.
The env vars are needed until go workspaces lands, then it can get
simpler.
Downsides to this:
1) If you don't call kube::golang::setup_env, it might work but will
just splat results somewhere
2) The resultant binaries are not in _output/bin but instead in the
phony GOPATH/bin (which setup_env puts in PATH)
The `make` rules which auto-generate some of our API stuff are
incredibly baroque, and hard to maintain. They were originally added on
the assumption that we would stop checking generated files into git.
Since then we have moved away from that goal, and the worst problems
with generated files have been resolved.
Reasons to kill this:
* It is slow on every build, as opposed to just being slow when running
the generators. It is even slow to calculate that there's nothing to
update.
* Most development work doesn't involve changing APIs.
* It only covers about half (or less) of the generated code, and making
it cover more would be even slower.
* Approximately 1 person knows how this all works.
* We have CI to make sure changes do not get merged without updating
this code.
* We have corner cases where this does the WRONG thing and tracking
those down is ugly and hard in perpetuity.
So this commit puts all the same logic that WAS in the
Makefile.generated_files into update-codegen.sh.
I do not love this script, especially WRT sub-packages, but I am trying
not to boil the ocean. I hope to follow up with some more cleanups over
time.
I have tested this manually and with the scripts and it still seems to
catch errors properly.
This includes a change to kube::util::read-array to make it not unset
variables and not over-write non-array variables.
