🚀 go-pugleaf

I post this comment on the new treatment of the variable i in the for
loop 'for (int i ...)' not so much to say that I disagree with the 
change (in theory, anyway, I think it was a correct decision) as a
warning about how much code this will break.  I was apprehensive 
from the outset when I first discovered the change, but the extent
of the problems to be faced as a result hadn't really hit home until
I undertook the task of building groff for Windows NT.  It was a
real eye-opener to see how much reliance there was on the assumption
that the variable's lifetime continues until the end of the scop
enclosing the 'for' loop.  The worst was tbl, whose source compiled
without error and linked fine, even produced reasonable-looking
output until you looked closely and found out spanned columns 
weren't being handled properly.  It took quite a bit of digging to
find the culprit.  It looks a lot simpler in this boiled-down
version, but the original function was lengthy and convoluted
enough to make spotting the problem non-trivial:

   ...
   int c = in.get();
   ...
   while (...) {
      for (int c = 0; c < ncolumns; c++) { 
         ...
      }
      for (c = 0; c < ncolumns; c++) { // wrong c!
         ...
      }
   }
   ...
   for (;;) {
      if (c == tab_char || c == '\n') // oops! c isn't what it's
                                      //  supposed to be ...

I'd say it's hard to fault the authors of the original code, who
were just following the advice frequently given to C++ programmers
to delay creation of a variable until as close to its first use as 
possible (of course, it would have helped if they had used a
different name for the second variable :->} ).

I expect this is just a small taste of similar fun to come.

--
/*----------------------------------------------------------------------*/
/* Bob Kline                                       Stream International */
/* bob_kline@stream.com               formerly Corporate Software, Inc. */
/* voice: (703) 522-0820 x-311                      fax: (703) 522-5407 */
/*----------------------------------------------------------------------*/
---
[ comp.std.c++ is moderated.  Submission address: std-c++@ncar.ucar.edu.
  Contact address: std-c++-request@ncar.ucar.edu.  The moderation policy
  is summarized in http://dogbert.lbl.gov/~matt/std-c++/policy.html. ]