diff options
Diffstat (limited to 'stdlib/camlinternalFormat.ml')
-rw-r--r-- | stdlib/camlinternalFormat.ml | 17 |
1 files changed, 14 insertions, 3 deletions
diff --git a/stdlib/camlinternalFormat.ml b/stdlib/camlinternalFormat.ml index 1c35d7809..acaf16b76 100644 --- a/stdlib/camlinternalFormat.ml +++ b/stdlib/camlinternalFormat.ml @@ -1421,11 +1421,22 @@ let fmt_ebb_of_string str = (_, _, e, f) fmt_ebb = fun pct_ind str_ind end_ind plus sharp space ign pad -> if str_ind = end_ind then unexpected_end_of_format end_ind; - match str.[str_ind] with - | '0' .. '9' -> + let parse_literal str_ind = let new_ind, prec = parse_positive str_ind end_ind 0 in parse_conversion pct_ind (new_ind + 1) end_ind plus sharp space ign pad - (Lit_precision prec) str.[new_ind] + (Lit_precision prec) str.[new_ind] in + match str.[str_ind] with + | '0' .. '9' -> parse_literal str_ind + | ('+' | '-') when legacy_behavior -> + (* Legacy mode would accept and ignore '+' or '-' before the + integer describing the desired precision; not that this + cannot happen for padding width, as '+' and '-' already have + a semantics there. + + That said, the idea (supported by this tweak) that width and + precision literals are "integer literals" in the OCaml sense is + still blatantly wrong, as 123_456 or 0xFF are rejected. *) + parse_literal (str_ind + 1) | '*' -> parse_after_precision pct_ind (str_ind + 1) end_ind plus sharp space ign pad Arg_precision |