fix: lone quote in skipped line no longer consumes rest of file

[LeonidasZhak]

Fixes tidyverse/readr#1577

Problem

When skip is used and a skipped line contains a lone (unclosed) quote character, find_next_non_quoted_newline sets in_quote = true and never resets it. This causes all subsequent newlines to be treated as embedded within a quoted field, making the skip consume the entire file.

# Before fix: returns empty tibble
write('"\na,b\n1,2', "quote.csv")
read_csv("quote.csv", skip = 1)
#> # A tibble: 0 × 0

Fix

In find_next_non_quoted_newline (src/utils.h), when EOF is reached while in_quote is still true (indicating an unclosed quote), fall back to finding the first newline from the original start position. This ensures a lone quote in a skipped line doesn't consume the rest of the file.

# After fix: returns expected data
write('"\na,b\n1,2', "quote.csv")
read_csv("quote.csv", skip = 1)
#> # A tibble: 1 × 2
#>       a     b
#>   <dbl> <dbl>
#> 1     1     2

Changes

• src/utils.h: Add fallback logic in find_next_non_quoted_newline for unclosed quotes
• tests/testthat/test-vroom.R: Add regression test

Testing

• New test passes
• All existing tests pass (1163 pass, 5 pre-existing Unicode path encoding failures on macOS)
• Properly quoted fields with embedded newlines still work correctly

[LeonidasZhak]

I am withdrawing this PR as part of a broader cleanup of an oversized automated contribution batch. I am sorry for the review noise and the pressure this may have put on maintainers. Going forward I will keep contributions to this project much more limited and higher-signal, ideally only one or two focused PRs at a time. Thank you for maintaining the project.