[bug #59622] grog: incorrectly infers man pages produced by pod2man

[G. Branden Robinson]

Update of bug #59622 (project groff):

                  Status:             In Progress => Fixed                  
             Open/Closed:                    Open => Closed                 
         Planned Release:                    None => 1.23.0                 

    _______________________________________________________

Follow-up Comment #4:


commit b0de53c923bfd77191157f6caacff984e8ca5e82
Author: G. Branden Robinson <g.branden.robinson@gmail.com>
Date:   Wed Jun 30 19:30:00 2021 +1000

    [grog]: Refactor input parsing.
    
    * src/utils/grog/grog.pl:
      - Add scalar `use_compatibility_mode` (see below).
      - Add list `request` to store the names of all requests recognized by
        groff so that they aren't confused with macro names.
      - Add scalars `have_seen_first_macro_call` (replaces
        `before_first_command`, but at global scope), `is_continued_line`
        and `logical_line`.  The latter two enable us to handle *roff input
        line continuation correctly.
    
      (process_arguments): Set `use_compatibility_mode` if `-C` option
      specified.
    
      (process_input): Refactor to greatly simplify, to not attempt to read
      the first line of an input file as a special case, and to avoid
      sending `do_line` an undefined argument (when EOF is reached).
    
      (do_first_line): Delete.
    
      (do_line): Rewrite the early stages of input parsing.
      - Concatenate continued input lines, setting `is_continued_line` and
        returning early as each one is seen, storing the accumulating input
        in `logical_line`.
      - Check the input line for the form of comment deposited by Perl's
        Pod::Man, which uses a highly accented dialect of man(7); if it's
        present, inflate `man_score` to compensate for the page-private `IX`
        macro it defines but which duplicates the name and function of a
        4.2BSD-era ms(7) extension that would otherwise deceives our scoring
        mechanism, because Pod::Man produces `IX` calls to metastatic
        excess.  (An alternative to this kludge is documented in comments:
        if a "standard" macro is redefined, we could delete it from the
        relevant lists and hashes.)
      - Strip *roff comments from input.
      - Normalize control lines; convert the no-break control character to
        the regular one and remove unnecessary white space.
      - Remove brace escapes.
      - Recognize two-character macro calls when not followed by white space
        in compatibility mode.
      - Drop logic that erroneously attempted to infer soelim(1) use from
        macro calls and request invocations.  The grog(1) and soelim(1) man
        pages now both explain why such an effort was misguided.
      - Recognize macro definitions created by .am and .am1 requests (not
        just .de and .de1).
      - Ignore all other *roff requests.
      - What remains must be a ("standard") macro call, so set
        `have_seen_first_macro_call`.
    
    * src/utils/grog/grog.1.man (Limitations): Document a further
      restriction: don't change the escape character, either.
    
    * src/utils/grog/tests/smoke-test.sh: Comment out pic-detection test on
      soelim(1).  The pic macro calls are guarded by roff control structures
      and only worked previously by accident because grog did not recognize
      *roff input line continuation, now it does and the illusion is
      dispelled.  (A reliable way to fool grog before and after my
      refactoring is now documented in its man page.)
    
    Fixes <https://savannah.gnu.org/bugs/?59622>.


    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?59622>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/