[TK-01182] 2020 Code Coverage #64
Conversation
| @@ -12,14 +12,15 @@ | |||
|
|
|||
| # can be any github ref | |||
| # branch, tag, commit, etc. | |||
| ref = "2020-02-27-rust-stable"; | |||
| ref = "1ca45a6f1899947e7fd0addc951858c110a391a5"; | |||
| # ^^ this is HEAD of 2020-02-27-rust-stable-2 as of 2020-04-09 (david.b) | |||
There was a problem hiding this comment.
I couldn't just push to 2020-02-27-rust-stable without breaking everyone who's currently based off develop -- by using the actual commit hash here, this won't be a problem in the future.
There was a problem hiding this comment.
Do you mean that this hash refers to a commit in holochain/holonix?
| @@ -38,6 +40,7 @@ fn malformed_toml_error_is_friendly() { | |||
| } | |||
|
|
|||
| #[test] | |||
| #[ignore] // (david.b) [D-01034] these take minutes to run in nix/CI - disable until fixed | |||
There was a problem hiding this comment.
Weirdly, I haven't had this problem running the tests outside of nix, but I did see it on Art's nix-shell. What do you make of this?
There was a problem hiding this comment.
I haven't looked into it - usually nix-specific problems have to do with nix's funky linking.
$ readelf -d target/debug/sx_types-66e1c5faaaa26893 | grep RUNPATH
0x000000000000001d (RUNPATH) Library runpath: [/nix/store/wac2lyrm69vihamdpx7dnh5n35df2kji-dev-shell/lib64:/nix/store/wac2lyrm69vihamdpx7dnh5n35df2kji-dev-shell/lib:/nix/store/pnd2kl27sag76h23wa5kl95a76n3k9i3-glibc-2.27/lib:/nix/store/4l35nqpaiwzhfafrpby1xf7kfik7ai7c-gcc-8.3.0-lib/lib]
There was a problem hiding this comment.
I approve these changes to the code. Nice work!
As we merge this though, I have two follow-up questions that maybe deserve to become new issues:
- I would like to hear from you @neonphog, what would it take to avoid the need for this hack and instead just have
kcovrespectCARGO_TARGET_DIR? - What would it take to prevent "tooling-rot" by ensuring that this script continues to work as we change things? Do we want to run this in CI? Maybe even as a separate job?
Totally agree - This is a frustration for me as well. But |
Yep, we can totally run this in CI - it'll be a little weird and not accomplish anything since we can't publish the results - I'll bring it up as a question in standup today. |
Cool. It would feel weird though to delay people getting their CI results as the cost for this. I wonder if ideally we would want a notification that appears on a PR, saying that it broke the tool, but not blocking the PR. Kind of like Rust has for PRs that break |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
We cannot do codecov.io right now unless we pay for it because we are using a private repo.
This PR just adds a
hc-coverage-testcommand that will pop the report in a browser. We can also make a circle/githubaction job that will block the build if we don't get > some coverage % if we want, but that's not included in this request.