[groff] 12/12: [troff]: Fix Savannah #64484.

[G. Branden Robinson]

gbranden pushed a commit to branch master
in repository groff.

commit cba317296a9e835e44a3792b52232dce81d53407
Author: G. Branden Robinson <g.branden.robinson@gmail.com>
AuthorDate: Tue Nov 19 18:55:54 2024 -0600

    [troff]: Fix Savannah #64484.
    
    This change restores groff 1.23.0's handling of characters in `device`
    request arguments, except that {1} special character escape sequences
    are interpreted (except for composite special characters, a planned
    future development) and {2} unprintable character codes such as ASCII 2
    are no longer emitted in GNU troff output by this means.
    
    * src/roff/troff/input.cpp (device_request): Refactor and revise
      argument interpretation.  Reconstitute an encoded unbreakable space
      code point as its escape sequence (spelled with the default escape
      character).  Throw warning in category "syntax" if a valid but
      unprintable input character is encountered.  Throw warning in category
      "syntax" if an escape sequence other than `\[` is encountered.  (A
      reconstituted `\~` is not warned about.)
    
    * src/roff/troff/troff.1.man (Warnings): Document additional
      circumstances under which warnings in "syntax" category are thrown.
    
    Fixes <https://savannah.gnu.org/bugs/?64484>.
---
 ChangeLog                  | 23 +++++++++++++++++++++++
 src/roff/troff/input.cpp   | 38 ++++++++++++++++++++++++++++++++------
 src/roff/troff/troff.1.man |  3 +++
 3 files changed, 58 insertions(+), 6 deletions(-)

diff --git a/ChangeLog b/ChangeLog
index 3bb59a237..03df62bba 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,26 @@
+2024-11-19  G. Branden Robinson <g.branden.robinson@gmail.com>
+
+       [troff]: Fix Savannah #64484.  This change restores groff
+       1.23.0's handling of characters in `device` request arguments,
+       except that {1} special character escape sequences are
+       interpreted (except for composite special characters, a planned
+       future development) and {2} unprintable character codes such as
+       ASCII 2 are no longer emitted in GNU troff output by this means.
+
+       * src/roff/troff/input.cpp (device_request): Refactor and revise
+       argument interpretation.  Reconstitute an encoded unbreakable
+       space code point as its escape sequence (spelled with the
+       default escape character).  Throw warning in category "syntax"
+       if a valid but unprintable input character is encountered.
+       Throw warning in category "syntax" if an escape sequence other
+       than `\[` is encountered.  (A reconstituted `\~` is not warned
+       about.)
+       * src/roff/troff/troff.1.man (Warnings): Document additional
+       circumstances under which warnings in "syntax" category are
+       thrown.
+
+       Fixes <https://savannah.gnu.org/bugs/?64484>.
+
 2024-11-19  G. Branden Robinson <g.branden.robinson@gmail.com>
 
        [groff]: Regression-test Savannah #64484.
diff --git a/src/roff/troff/input.cpp b/src/roff/troff/input.cpp
index 8fafa8a0a..0aeee3760 100644
--- a/src/roff/troff/input.cpp
+++ b/src/roff/troff/input.cpp
@@ -6013,17 +6013,43 @@ static void device_request()
   }
   if ((curdiv == topdiv) && (topdiv->before_first_page_status > 0))
     topdiv->begin_page();
-  // Null characters can correspond to node types like vmotion_node that
-  // are unrepresentable in a device extension command, and got scrubbed
-  // by `asciify`.
-  for (int prevc = c; (c != '\0') && (c != '\n') && (c != EOF);
+  for (;
+      (c != '\0') && (c != '\n') && (c != EOF);
        c = get_copy(0 /* nullptr */)) {
-    if (!('\\' == c) && (prevc != '\\'))
+    // We may encounter some of the C0 and C1 character codes GNU troff
+    // uses for special purposes; see src/roff/troff/input.h.  They
+    // produce nothing in grout.  Warn only about the ones that are left
+    // for the user's purposes.  Use octal because input.h does.  Ignore
+    // 8-bit codes in general.  grout is an ISO 646 file format.
+    if (ESCAPE_TILDE == c) {
+      mac.append('\\');
+      mac.append('~');
+    }
+    else if ((c < 015) || (c >= 0177))
+      warning (WARN_SYNTAX, "ignoring character code %1 in device"
+              " extension command request argument", c);
+    else if (c != '\\')
       mac.append(c);
     else {
       int c1 = get_copy(0 /* nullptr */);
-      if (c1 != '[')
+      if (c1 != '[') {
+       mac.append(c);
        mac.append(c1);
+       string chardesc = "";
+       if (csprint(c1)) {
+         chardesc += "'";
+         chardesc += char(c1);
+         chardesc += "'";
+       }
+       else {
+         chardesc += "character code ";
+         chardesc += i_to_a(c1);
+       }
+       chardesc += '\0'; // make it safe for .contents()
+       warning (WARN_SYNTAX, "not interpreting escaped %1 in device"
+                  " extension command request argument",
+                  chardesc.contents());
+      }
       else {
        // Does the input resemble a valid (bracket-form) special
        // character escape sequence?
diff --git a/src/roff/troff/troff.1.man b/src/roff/troff/troff.1.man
index 5a633658e..b75028493 100644
--- a/src/roff/troff/troff.1.man
+++ b/src/roff/troff/troff.1.man
@@ -755,6 +755,9 @@ It never occurs in compatibility mode.
 A self-contradictory hyphenation mode was requested;
 an empty or incomplete numeric expression was encountered;
 an operand to a numeric operator was missing;
+a recognized but inapposite escape sequence
+or unprintable character code
+was used in a device extension command request;
 an attempt was made to define a recursive,
 empty,
 or nonsensical character class;