Michael Bahr
Michael Bahr
Some more things I tested: - In my previous version with client/playwright, I didn't use chdir on the target. I attempted running the e2e target without chdir in https://github.com/sourcegraph/sourcegraph/commit/0d3a28b9b407804d01b91d8dcdd565d8d05750e7.
New insights! I think that [this issue](https://github.com/sourcegraph/sourcegraph/issues/60881#issuecomment-2023011320) is not actually caused by multiple versions of playwright, but by a combination of how playwright, bazel and node work. The culprit here...
I tried three variations of the approach to use js_test instead with an entrypoint. ## node_modules entry with playwright dependency ``` js_test( name = "e2e_test", args = [ "test", "--config...
@jamesmcnamara Very happy to see your progress! I can confirm that without importing `:web-sveltekit` into the `e2e_test` target I get the same errors that you describe. I tried to dive...
@jamesmcnamara Just wanted to check in if you've been able to make more progress. Since I'm not able to make progress myself I want to be supportive with whatever approach...
I'm continuing to track this ticket. My current status is that things have been set up so that we can work with Aspect.
I believe that the currently perceived slowness comes from backfills sequentially building jobs, then running jobs, and finally persisting the results. This means that all jobs need to complete before...
Here's what users see while the insight is still fetching historical data: https://github.com/sourcegraph/sourcegraph/blob/a659fbbb083fa46d8518a798443570a7f23bb48b/client/web/src/enterprise/insights/components/insights-view-grid/components/backend-insight/components/backend-insight-alerts/BackendInsightAlerts.tsx#L36-L47
I found a test case that takes long enough that we can measure improvements against. Build a code insight for all repositories with this query: `(.+)\sv(?:[^\s]+)(?: // indirect)? file:go\.mod patterntype:regexp...
> wouldn't there potentially be a different reason per incomplete repository? In theory yes. Currently we have only timeout and generic as reasons. From what I can see the code...