[bug #64957] [troff] commit 429723c3ec broke `device` and `output` requests, and `pdfbackground`

[G. Branden Robinson]

URL:
  <https://savannah.gnu.org/bugs/?64957>

                 Summary: [troff] commit 429723c3ec broke `device` and
`output` requests, and `pdfbackground`
                   Group: GNU roff
               Submitter: gbranden
               Submitted: Fri 01 Dec 2023 07:44:09 PM UTC
                Category: Core
                Severity: 3 - Normal
              Item Group: Incorrect behaviour
                  Status: In Progress
                 Privacy: Public
             Assigned to: gbranden
             Open/Closed: Open
         Discussion Lock: Any
         Planned Release: None


    _______________________________________________________

Follow-up Comments:


-------------------------------------------------------
Date: Fri 01 Dec 2023 07:44:09 PM UTC By: G. Branden Robinson <gbranden>
This problem
[https://git.savannah.gnu.org/cgit/groff.git/commit/?id=429723c3ec773a36313001ba70862fa6a7408984
arose on 10 November], but took a little while to notice.

The GNU _troff_(1) internal function `has_arg()` has a side effect of
advancing the token pointer (a sort of cursor into the input stream).  So when
I changed `device_request()` and `output_request()` to use `has_arg()`, the
first character of the argument got stripped, making it unintelligible to the
output driver (_gropdf_(1)).

In other words,


.device pdf: niftyfeature arg1 arg2 arg3


became this.


.device df: niftyfeature arg1 arg2 arg3


The visible effect of this is that the background colors and box borders in
_msboxes.pdf_ went missing.

Nothing else broke (as far as I can tell) because (A) pretty much no part of
_groff_ appears to use the `output` request and (B) in most places where
device control commands are issued, it is done with `\X` escape sequences
which are not effected.

This isn't a problem with most other users of `has_arg()` because they proceed
to _use_ the token pointer.  But these functions don't, because they gather
input with `get_copy()` instead.

I have a patch in preparation.







    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?64957>

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