Fix longitudinal report #5783
Fix longitudinal report #5783
Conversation
Unit Test Results 18 files + 6 18 suites +6 9h 39m 9s ⏱️ + 3h 12m 7s For more details on these failures and errors, see this check. Results for commit 78779bd. ± Comparison against base commit 65f4151. ♻️ This comment has been updated with latest results. |
| @@ -172,7 +173,7 @@ def dedup(group): | |||
| for w in workflows | |||
| if ( | |||
| pandas.to_datetime(w["created_at"]) | |||
| > pandas.Timestamp.now(tz="UTC") - pandas.Timedelta(days=90) | |||
| > pandas.Timestamp.now(tz="UTC") - pandas.Timedelta(days=60) | |||
There was a problem hiding this comment.
I'm starting to be concerned about bumping up against GitHub's rate limit for the use of GITHUB_TOKEN in actions (currently 1000 requests per hour). Roughly speaking, we are downloading ~10 artifacts per workflow run, plus a slew of administrative requests here. An alternative would be to just choose the most recent ~60-80 runs or thereabouts.
There was a problem hiding this comment.
ok, got it.
I think what we're looking for is indeed "last X runs" and not "last X days". The way I understand this code, that would be very easy to change, wouldn't it?
I don't think there is an easy way -- the test failure in question was checking for a specific string in the HTML for the dask dashboard. To be clear: this fix shouldn't be necessary. Vega specs should be robust to unusual strings, and indeed, this probably represents a security hole in Vega where a malicious user could inject HTML/JS to do something bad to a user. I suspect a bug was recently introduced in the Altair+Vega+VegaLite stack somewhere, but I wasn't able to find it in a timely fashion, and I wanted to unblock this visualization. |
| @@ -172,7 +173,7 @@ def dedup(group): | |||
| for w in workflows | |||
| if ( | |||
| pandas.to_datetime(w["created_at"]) | |||
| > pandas.Timestamp.now(tz="UTC") - pandas.Timedelta(days=90) | |||
| > pandas.Timestamp.now(tz="UTC") - pandas.Timedelta(days=60) | |||
There was a problem hiding this comment.
ok, got it.
I think what we're looking for is indeed "last X runs" and not "last X days". The way I understand this code, that would be very easy to change, wouldn't it?
|
As always, thanks @ian-r-rose for fixing these builds so quickly |
That's right -- I'll push a change for this right now |
ian-r-rose
force-pushed
the
test-report
branch
from
3687a83
to
78779bd
Compare
|
Done |
There was some weird variation in behavior here depending on precise versions of libraries that I didn't really track down, but the heart of the issue was that, a few days ago, a test failure message included a full HTML string that was not escaped, and messed up the resulting HTML document.
Fix is live on my fork
pre-commit run --all-files