fix(trace): Fixed infinite line-wrapping #529
Conversation
There was a fun bug where if the offset in the line-wrapping was ever 0, it would never force a hard-wrap at the limit and also set the line back to itself in the slice. As such it would just hopelessly spin forever trying its best to wrap a line. This bug would also happen if the character that landed *right* on the end of the wrapping boundary was whitespace, causing the offset to be 0.
|
Thank you for looking at this! We could probably also write some tests for this stuff (running on the development system with |
|
Yeah no problem! I was ahem borrowing a big chunk of the code for a thing I'm working on when I found the bug :v Some proper tests would be nice, but alas, the spoons and effort needed is not in the budget right now, for me at least. As for the edge cases I found, there are some very rare cases where line wrapping just goes mental, I suspect there is a subtle off-by-one somewhere, where it imagines that it can to a soft-wrap right after the first character? This also seemes very inconsistent, as most of the time it would be fine, but then sometimes just, has a conniption
I've just given up for what i'm doing and force a hard-wrap at the boundary every time, but fixing it would be nice, lol |
|
oh! i ended up making a similar change in #551, but:
totally missed that. that bug's still there. |
|
I think I fixed this on my end months ago and just forgor to update this PR, I can do so and update this PR or just drop this one and it can be done elsewhere, i'm good with whatever, just let me know |
3 participants
Within
Writer::write_strthere is a bug in the line wrapping logic, where if theoffsetis ever0, it would never force a hard-wrap and cause the poor line wrapping code to spin forever, hopelessly trying its best, but making no progress.There are 2 cases where
offsetcan be0, the first is when it can't find any whitespace to wrap on, and the second being when the very last character of the wrap boundary was whitespace.This fixes the primary problem by always forcing an ugly hard-wrap at the boundary if
offsetis ever0, and updating theoffsetvalue to be that boundary limit. This then allows the sub-slice of the line to work as expected and not run in circles forever.EDIT: Turns out there are still a few edge cases, i'll keep poking it :v