-
Notifications
You must be signed in to change notification settings - Fork 171
Possible closedness regression in v0.3.0-beta.1 #642
Comments
@addreas thanks for the detailed report with small reproducer. That is extremely helpful. BTW, We would be very interested to investigate repos that take 32.1s. I don't think there should be many repos that take more than 1s or 2s. We reduced several hour-long running configs to less than 1s or 2s, so it could very well be that yours is also is suffering from some yet-to-be-fixed performance bug. We are aware of a few more potential issues, but we tend to prioritize the ones that we know are causing issues for users. @myitcv is working on a setup where we can validate new versions again external repos that users can register. So maybe waiting for that makes sense, if you're interested for us to take a look. |
Would in some cases recursively close. Now uses just one method for marking closedness. Marking Closed in the Vertex was only used so that the debug printer would show the closed status. This has now fixed. This was not possible before but is now as the code was recently simplified. Fixes #642 Change-Id: I9aab6a02f36ddbc8ed9bce356a2f4ad77cd30cda
Validating against external repos seems like a great idea so I took some time to setup a repository over here that is a bit smaller but seems to show the same performance characteristics. Wouldn't be surprised if you are correct and I'm doing something weird. Always had a feeling that performance would be ironed out at some point, so I never thought too much about it. Also found another issue while setting that up, but I just saw the beta2 release so perhaps I don't have to. Continually impressed with the development pace of this project 🥇 |
@addreas - just following up on the point about regression testing, |
This issue has been migrated to cue-lang/cue#642. For more details about CUE's migration to a new home, please see cue-lang/cue#1078. |
What version of CUE are you using (
cue version
)?Git tags
v0.3.0-alpha6
andv0.3.0-beta.1
, some bisecting points to cea55b2 as the culpritDoes this issue reproduce with the latest release?
Yes
What did you do?
Got it down to this:
What did you expect to see?
For it to type check correctly
What did you see instead?
Was easily worked around, and ended up being a lot less convoluted. By the way, a
cue vet
of my entire repository dropped from 55.5s to 32.1s, great work on the performance improvements! 👍The text was updated successfully, but these errors were encountered: