Skip to content

Hold the two large text bugs a reader reported - #597

Merged
andiwand merged 1 commit into
mainfrom
test/large-text-regressions
Aug 18, 2026
Merged

Hold the two large text bugs a reader reported#597
andiwand merged 1 commit into
mainfrom
test/large-text-regressions

Conversation

@andiwand

@andiwand andiwand commented Aug 17, 2026

Copy link
Copy Markdown
Member

From user feedback: "Funktioniert seit Neuestem nicht mehr brauchbar mit größeren TXT Files (ca. 1 MB): File laden dauert lange, ebenso die unmögliche Textstellensuche! War vorher prima."

Two tests, one per bug. Against the shipped odrcore 6.7.1 they fail with exactly the reported symptoms:

opening and searching 6811 lines took 29150ms, over the 15000ms budget
prose was read as a table expected:<text/[plain]> but was:<text/[csv]>

Both go green once the app takes an odrcore carrying:

Why here and not only in the engine

Both engine PRs carry their own tests. What neither can say is that a megabyte of text opens in this app and is then searchable — the loader, the cache, the WebView and Find are all on this side of the boundary. aMegabyteOfTextOpensAndIsSearchable asserts the search finds every line and that opening plus searching fits in a budget; proseWithCommasStaysText asserts the mime type the app ends up showing.

The budget is 15s against a measured 0.26s for layout after the fix, so it is a ceiling a broken render cannot fit under rather than a benchmark.

The text file is generated rather than checked in — a megabyte of filler is not worth a git object, and nothing here depends on which words it is.

🤖 Generated with Claude Code

@andiwand
andiwand force-pushed the test/large-text-regressions branch from c559116 to 707ced5 Compare August 18, 2026 07:04
@andiwand
andiwand marked this pull request as ready for review August 18, 2026 07:14
Both fail against odrcore 6.7.1 and pass once the fixes for them are taken:

    took 29150ms, over the 15000ms budget
    expected:<text/[plain]> but was:<text/[csv]>

They are end to end on purpose. The engine's own tests cover the detection
rule and the layout it does; what neither can say is that a megabyte of text
opens in this app and can then be searched.

The file is generated rather than checked in - a megabyte of filler is not
worth a git object, and nothing here depends on which words it is.

Do not merge before the odrcore bump.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TouAQNfktsp9THcceennEX
@andiwand
andiwand force-pushed the test/large-text-regressions branch from 707ced5 to 6d5036a Compare August 18, 2026 08:21
@andiwand
andiwand merged commit 81e16cd into main Aug 18, 2026
7 checks passed
@andiwand
andiwand deleted the test/large-text-regressions branch August 18, 2026 08:46
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant