segfault with read_fwf() and fwf_positions()

[MatthieuStigler]

I get a segfault error when running the code below, tested on two different Ubuntu machines. vroom (1.6.5) and readr (2.1.5) seem to be at the latest version currently available.

library(readr)
packageVersion("readr")
packageVersion("vroom")
starts <- c(0, 13, 19, 26)
ends <- c(11, 17, 24, NA)


txt <- "ID           Arch  Saved  What
C014KLCJSHY  -     -      #string
C014QHAKJHY  -     -      #somethin"
cat(txt)

channel_ids <- read_fwf(
  file=  txt,
  col_positions = fwf_positions(starts, ends))

Will give:

*** caught segfault ***
address 0x7f6c910fbfff, cause 'memory not mapped'
Traceback:
1: vroom_fwf_(file, as.integer(col_positions$begin), as.integer(col_positions$end), trim_ws = trim_ws, col_names = col_positions$col_names, col_types = col_types, col_select = col_select, name_repair = .name_repair, id = id, na = na, guess_max = guess_max, skip = skip, comment = comment, skip_empty_rows = skip_empty_rows, n_max = n_max, num_threads = num_threads, altrep = vroom_altrep(altrep), locale = locale, progress = progress)
2: vroom::vroom_fwf(file, col_positions = col_positions, col_types = col_types, col_select = { { col_select } }, id = id, .name_repair = name_repair, locale = locale, na = na, comment = comment, skip_empty_rows = skip_empty_rows, trim_ws = trim_ws, skip = skip, n_max = n_max, guess_max = guess_max, show_col_types = show_col_types, progress = progress, altrep = lazy, num_threads = num_threads)
3: read_fwf(file = txt, col_positions = fwf_positions(starts, ends))
Possible actions:
1: abort (with core dump, if enabled)

[matsokvist]

I noticed that fwf_positions() subtracts 1 from the start positions I got from fwf_empty() , leading to R crashing when using them in read_fwf().

With this change the above example runs:

col_positions = fwf_positions(starts**+1**, ends)

Alternatively, fwf_empty() reproduces the starts and ends vectors given in the example, and using it directly also works:

col_positions = fwf_empty(txt)

Dont't know what the idea behind the difference in output of the two functions is, but it was not what I expected....