Repository navigation
Record a snapshot of the pages that never opened a browser - #3
Merged
Merged
Conversation
The tier-0 result carried neither the capture nor the served bytes, so -snapshot on a page answered without a browser wrote a file containing little more than a URL, and replay rejected it with "contains no capture" -- which reads like a corrupt file rather than one that was never given anything to hold. Every other return path in Distill carried both; this one, the path most pages take, was the exception. The consequence is worse than a missing feature. Snapshots exist so a bug can be reproduced offline against a user's attached file, and the bugs that were cheapest to reproduce were exactly the ones that could not be attached: every text-welding and reading-order defect fixed in this branch was a tier-0 defect, found by re-running live sites and fighting the network for it. So Read no longer refuses a snapshot that has the served HTML but no capture, and Replay re-runs the static extraction over those bytes to rebuild the same input the live run used. go.dev now records twenty-three kilobytes instead of a hundred and forty-three bytes, and replays to an identical content hash: same ninety-one blocks, same six sections, same hundred and twenty-four actions. A snapshot carrying neither a capture nor any HTML is still refused, and says which of the two it is missing. Found by running every command over the corpus rather than distill alone.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The tier-0 result carried neither the capture nor the served bytes, so -snapshot on a page answered without a browser wrote a file containing little more than a URL, and replay rejected it with "contains no capture" -- which reads like a corrupt file rather than one that was never given anything to hold. Every other return path in Distill carried both; this one, the path most pages take, was the exception.
The consequence is worse than a missing feature. Snapshots exist so a bug can be reproduced offline against a user's attached file, and the bugs that were cheapest to reproduce were exactly the ones that could not be attached: every text-welding and reading-order defect fixed in this branch was a tier-0 defect, found by re-running live sites and fighting the network for it.
So Read no longer refuses a snapshot that has the served HTML but no capture, and Replay re-runs the static extraction over those bytes to rebuild the same input the live run used. go.dev now records twenty-three kilobytes instead of a hundred and forty-three bytes, and replays to an identical content hash: same ninety-one blocks, same six sections, same hundred and twenty-four actions.
A snapshot carrying neither a capture nor any HTML is still refused, and says which of the two it is missing.
Found by running every command over the corpus rather than distill alone.