add skip_failed flag #121
add skip_failed flag #121
Conversation
| @@ -47,7 +47,7 @@ jobs: | |||
| # Use the community bundle because it's smaller and faster | |||
| git clone --recurse-submodules https://github.com/adafruit/CircuitPython_Community_Bundle.git | |||
| cd CircuitPython_Community_Bundle | |||
| circuitpython-build-bundles --filename_prefix test-bundle --library_location libraries --library_depth 2 | |||
| circuitpython-build-bundles --filename_prefix test-bundle --library_location libraries --library_depth 2 --skip_failed | |||
There was a problem hiding this comment.
I added more notes to the related issue but I want to note here that I'm uneasy about adding skip_failed during the test of circuitpython-build-tools: If a bug is introduced into the code here, --skip_failed might cover it up.
Instead, consider using the communty bundle at a specific ref/tag that is known to build properly, and remove --skip_failed.
There was a problem hiding this comment.
What do you think about having this task use the main Adafruit Bundle instead of the community bundle?
The comment here notes that community bundle was chosen because it's smaller and thus faster. It is still true that the community bundle is smaller, but it's also gaining ground at ~158 libraries right now. I think the main bundle is ~340ish but am not certain. I imagine that the disparity was greater when this task was initially set up.
I am not so sure that the increased speed is that big of a benefit. As I understand it this is always run automatically inside of github actions, there isn't typically a person sitting and waiting for it to complete.
In my mind pointing it to the main Adafruit Bundle would also be nice because it will give another chance to discover ASAP if any changes to the build-tool could cause any issues for the Adafruit libraries or bundle.
|
Closing this in favor of: adafruit/adabot#378 I do still think it would make sense for the actions to test against the Adafruit Bundle instead of the community one. But that could be done separately and doesn't need this skip_failed concept to exist. |
Resolves: #120
This adds a new
--skip_failedflag to the CLI and relevant python functions, it defaults toFalseif unused so behavior stays the same as today.If the new flag is passed it will continue the build process instead of bailing after a failed build within a library.
I will open a PR shortly in the community bundle that will utilize this flag to allow it build what it can even if some libraries do fail.
An open question that I'm interested in thoughts on is what to do in the case of the failed builds. It will still print the errors in the actions output, but with the Task passing it's unlikely to be seen there by anyone. I would guess without some other mechanism the most likely time it would be discovered is when someone tries to find that library within the bundle and is unable to.
Maybe a notice message or warning message