Repository navigation
Commit bdfde5a
committed
🐛 Assert #870's failure row against the reason it recorded, not a phrase
UI13 asserted the drawn reason contained "Unexpected reserved word". That is V8
speaking: Deno compiles an eval block itself, while Node and Bun reach it through
esbuild and say "Transform failed with 1 error". The fixture is the same real
compiler failure in all three, so the row was green on the runtime it was written
on and red on the two it was not — `test-node (2/10)` and `test-bun (1/5)`.
The row now compares what is drawn with what this run actually retained: the
displayed reason has to be the beginning of the recorded message, flattened to
one line and cut to its region. The expected answer is derived here rather than by
calling the production formatter, so the row still fails if that formatter stops
normalizing or stops bounding.
Everything else it claimed is untouched: the one-line rule, the region bound, the
three-size rendered geometry, that it is the message and not the serialized record
around it, and that a cold projection answers with the identical row.
Omitting the reason reddens it, and so does replacing it with unrelated text.1 parent df98bf0 commit bdfde5a
1 file changed
Lines changed: 10 additions & 3 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
2969 | 2969 | | |
2970 | 2970 | | |
2971 | 2971 | | |
2972 | | - | |
2973 | | - | |
2974 | | - | |
| 2972 | + | |
| 2973 | + | |
| 2974 | + | |
| 2975 | + | |
| 2976 | + | |
| 2977 | + | |
| 2978 | + | |
| 2979 | + | |
| 2980 | + | |
| 2981 | + | |
2975 | 2982 | | |
2976 | 2983 | | |
2977 | 2984 | | |
| |||
0 commit comments