JdBP>>         if (FILE * f = fopen("file.txt","r")) {
 JdBP>>             // ...
 JdBP>>             fclose(f) ;
 JdBP>>         }
 DM> Well, I would *hope* that whoever was bright enough to note that this
 DM> was C++-only code, would also note that the implicit check against
 DM> NULL makes this marginally confusing, and one would probably be
 DM> better off moving the declaration (and initialization) of f to before
 DM> the if anyway.
I disagree.  Although I've seen a few "style guides" recommending to always 
make the test of a pointer against NULL explicit, I think that this is a 
mistake, founded in the belief that without the "== NULL" at the end, the 
condition of an if statement means something else.  But it doesn't mean 
something else, and "== NULL" is really superfluous.  So I don't think that 
it's a valid criticism, since the "confusion" doesn't exist once one has 
passed the stage of C++ novice, and one knows the rules for converting 
pointers to bools.
 DM> FILE* f = fopen("file.txt","r");
 DM> if (NULL != f) 
 DM> {
 DM>       // ...
 DM>       fclose(f);
 DM> }
 DM> 
 DM> Now this is much more readable (IMNSHO), and, as a bonus, legal 
 DM> C under the proposed C9X.  :-)
Which completely misses the point that the syntax in the original example has 
a much smaller scope for `f' and is a very useful idiom because of this.
 ¯ JdeBP ®
--- FleetStreet 1.19 NR
---------------