Unpin coverage #3384

New Issue

sajith · 2020-08-18T20:13:12Z

sajith commented

2020-08-18 20:13:12 +00:00

#3267 pinned some dependencies including coverage, because integration tests were failing on CI. It isn't clear if coverage still needs to remain pinned at ~= 4.5.

Let us unpin coverage and see what happens in CI!

For more context, the follow-up ticket to this would be #3385, which will attempt to use coveralls.io for code coverage checks. We are not entirely satisfied with codecov.io, which we currently use. For example, it is not always clear why codecov checks turn red/disapprove of certain PRs, even when test coverage remains unchanged.

Now, in order to upload coverage reports to codecov.io, we will need to use coveralls-python, which requires that coverage reports should be in coverage 5.0 format. Which is the real motivation for unpinning coverage.

If coveralls.io works better than codecov.io for us, we can switch to the former. Even if it doesn't, this change does not affect codecov: codecov can continue working with the reports we submit in XML format.

(See coverage changelog: "Coverage’s data storage has changed. In version 4.x, .coverage files were basically JSON. Now, they are SQLite databases.")

#3267 pinned some dependencies including coverage, because integration tests were failing on CI. It isn't clear if coverage still needs to remain pinned at ~= 4.5. Let us unpin coverage and see what happens in CI! For more context, the follow-up ticket to this would be #3385, which will attempt to use [coveralls.io](https://coveralls.io) for code coverage checks. We are not entirely satisfied with [codecov.io](https://codecov.io), which we currently use. For example, it is not always clear why codecov checks turn red/disapprove of certain PRs, even when test coverage remains unchanged. Now, in order to upload coverage reports to codecov.io, we will need to use [coveralls-python](https://github.com/coveralls-clients/coveralls-python), which requires that coverage reports should be in coverage 5.0 format. Which is the real motivation for unpinning coverage. *If* coveralls.io works better than codecov.io for us, we can switch to the former. Even if it doesn't, this change does not affect codecov: codecov can continue working with the reports we submit in XML format. (See coverage [changelog](https://coverage.readthedocs.io/en/coverage-5.2.1/changes.html#version-5-0a2-2018-09-03): "Coverage’s data storage has changed. In version 4.x, .coverage files were basically JSON. Now, they are SQLite databases.")

sajith added the

labels 2020-08-18 20:13:12 +00:00

sajith added this to the undecided milestone 2020-08-18 20:13:12 +00:00

sajith self-assigned this 2020-08-19 15:33:16 +00:00

sajith was unassigned by exarkun

2020-12-29 18:23:05 +00:00

exarkun self-assigned this 2020-12-29 18:23:05 +00:00

GitHub <noreply@github.com> commented

2020-12-30 01:36:16 +00:00

In 1bdca90/trunk:

Merge pull request #950 from LeastAuthority/3384.unpin-coverage

Update coverage pin

Fixes: ticket:3384

In [1bdca90/trunk](/tahoe-lafs/trac-2024-07-25/commit/1bdca909faf7fd6f0f27311c9cfbf1d127cf1e0b): ``` Merge pull request #950 from LeastAuthority/3384.unpin-coverage Update coverage pin Fixes: ticket:3384 ```

tahoe-lafs added the

fixed

label 2020-12-30 01:36:16 +00:00

GitHub <noreply@github.com> closed this issue

2020-12-30 01:36:16 +00:00

Sign in to join this conversation.