[Bug fortran/112586] New: [F2023] Add "A statement shall not have more than one million characters." warning

[Bug fortran/112586] New: [F2023] Add "A statement shall not have more than one million characters." warning

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112586

            Bug ID: 112586
           Summary: [F2023] Add "A statement shall not have more than one
                    million characters." warning
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: diagnostic
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: burnus at gcc dot gnu.org
  Target Milestone: ---

See also https://gcc.gnu.org/pipermail/gcc-patches/2023-November/636868.html

GCC currently has a warning for:

              if (gfc_notification_std (GFC_STD_GNU) || pedantic)
                gfc_warning (0, "Limit of %d continuations exceeded in "
                             "statement at %C",

for both fixed and free form source code. This matches:

Fortran 2018's
  "A statement shall not have more than 255 continuation lines."


But Fortran 2023 now has:
  "A statement shall not have more than one million characters."

Where a 'statement' is "A Fortran statement is a sequence of one or more
complete or partial lines." - and a 'line' is: "sequence of zero or more
characters".


It is not completely clear to me in how far comments count; It seems as if
comment lines (blank lines or blanks followed by a comment) do not count - but
do in-line comments count? What about '&'?  Does a (partial) line start with
the first non-blank character or already at the first column? It probably ends
with '\n' and ';' and neither counts.

* * *

The parsing is done in scanner.cc's gfc_next_char_literal — as code is
reparsed,  i.e. gfc_current_locus can move to older code, we cannot simply
count the number of characters at "return c".

BTW: As scanner.cc currently does not handle ';', which had to be added in that
case.

Actually, it is not quite clear to me how the counting of continuation lines
works with 'gfc_current_locus = old_loc' handling. As long we only jump back in
the same line, it's fine - but I expect that reparsing might cause problems
with the count handling. This count is only used by warning above, but
nonetheless.