[groff] 06/06: src/roff/troff/input.cpp: Report bad digits.

[G. Branden Robinson]

gbranden pushed a commit to branch master
in repository groff.

commit f23fb69b7b1c6d30c2a3ca12ea871f42e4e3a61a
Author: G. Branden Robinson <address@hidden>
AuthorDate: Tue Mar 31 21:59:25 2020 +1100

    src/roff/troff/input.cpp: Report bad digits.
    
    (read_size): Improve bad digit diagnostic by reporting what the bad
    digit actually is; literally if it is a printable character, and by
    numerical code otherwise.  Also replace repeated two-line idiom with no
    stream-validity error-checking with call to...
    
    (read_size_next_byte): Add helper function to perform EOF and newline
    checking every time read_size() takes a byte from the input stream;
    issue a diagnostic if either is found.
---
 ChangeLog                | 11 ++++++++++
 src/roff/troff/input.cpp | 56 ++++++++++++++++++++++++++++++++----------------
 2 files changed, 48 insertions(+), 19 deletions(-)

diff --git a/ChangeLog b/ChangeLog
index 2becdf0..0281062 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,14 @@
+2020-03-31  G. Branden Robinson <address@hidden>
+
+       * src/roff/troff/input.cpp (read_size): Improve bad digit
+       diagnostic by reporting what the bad digit actually is;
+       literally if it is a printable character, and by numerical code
+       otherwise.  Also replace repeated two-line idiom with no
+       stream-validity error-checking with call to...
+       (read_size_next_byte): Add helper function to perform EOF and
+       newline checking every time read_size() takes a byte from the
+       input stream; issue a diagnostic if either is found.
+
 2020-02-10  G. Branden Robinson <address@hidden>
 
        * doc/groff.texi:
diff --git a/src/roff/troff/input.cpp b/src/roff/troff/input.cpp
index 8a642ba..3231933 100644
--- a/src/roff/troff/input.cpp
+++ b/src/roff/troff/input.cpp
@@ -5027,45 +5027,60 @@ static int get_line_arg(units *n, unsigned char si, 
charinfo **cp)
   return 0;
 }
 
+// In groff prior to 1.22.4, read_size() (below) had an idiom of doing:
+//   tok.next();
+//   c = tok.ch();
+// ...without checking the token for EOF or newline.  Let's do better.
+static int read_size_next_byte(void)
+{
+  int saved_lineno;
+  const char *saved_filename;
+  // Reading the next token could discard the filename (EOF) or change
+  // the line number (newline); save them.
+  if (!input_stack::get_location(0, &saved_filename, &saved_lineno))
+    // This should never happen.
+    fatal("I don't know where I am!");
+  tok.next();
+  int eof = tok.eof();
+  if (eof || tok.newline())
+      error_with_file_and_line(saved_filename, saved_lineno,
+                               "expected digit in point size; got %1",
+                               eof ? "end of file" : "newline");
+  return tok.ch();
+}
+
 static int read_size(int *x)
 {
-  tok.next();
-  int c = tok.ch();
+  int c = read_size_next_byte();
   int inc = 0;
   if (c == '-') {
     inc = -1;
-    tok.next();
-    c = tok.ch();
+    c = read_size_next_byte();
   }
   else if (c == '+') {
     inc = 1;
-    tok.next();
-    c = tok.ch();
+    c = read_size_next_byte();
   }
   int val = 0;         // pacify compiler
   int bad = 0;
   if (c == '(') {
-    tok.next();
-    c = tok.ch();
+    c = read_size_next_byte();
     if (!inc) {
       // allow an increment either before or after the left parenthesis
       if (c == '-') {
        inc = -1;
-       tok.next();
-       c = tok.ch();
+       c = read_size_next_byte();
       }
       else if (c == '+') {
        inc = 1;
-       tok.next();
-       c = tok.ch();
+       c = read_size_next_byte();
       }
     }
     if (!csdigit(c))
       bad = 1;
     else {
       val = c - '0';
-      tok.next();
-      c = tok.ch();
+      c = read_size_next_byte();
       if (!csdigit(c))
        bad = 1;
       else {
@@ -5077,8 +5092,7 @@ static int read_size(int *x)
   else if (csdigit(c)) {
     val = c - '0';
     if (!inc && c != '0' && c < '4') {
-      tok.next();
-      c = tok.ch();
+      c = read_size_next_byte();
       if (!csdigit(c))
        bad = 1;
       else
@@ -5090,8 +5104,7 @@ static int read_size(int *x)
     return 0;
   else {
     token start(tok);
-    tok.next();
-    c = tok.ch();
+    c = read_size_next_byte();
     if (!inc && (c == '-' || c == '+')) {
       inc = c == '+' ? 1 : -1;
       tok.next();
@@ -5133,7 +5146,12 @@ static int read_size(int *x)
     return 1;
   }
   else {
-    error("bad digit in point size");
+    // read_size_next_byte() already threw an error on EOF or newline.
+    if (!tok.eof() && !tok.newline())
+      if (csprint(c))
+        error("bad digit '%1' in point size", (char) c);
+      else
+        error("bad digit (character code %1) in point size", c);
     return 0;
   }
 }