[cups.general] Fwd: CUPS STR #1036 updated

[Michael Sweet]

Marc Balmer wrote:
> ...
> You are wrong in assuming that this is a compiler only warning. Your 
> code behaves different on plattforms where char is unsigned (like 
> macppc/OpenBSD).

We have access to systems that use both signed and unsigned chars
by default, and neither type of platform has a problem with the
code.

The type of check you references in STR #1036 was:

     if ((*ptr >= 0 && *ptr <= ' ') || *ptr == 127 || *ptr == '/')

On a system that has unsigned characters, this reduces to:

     if ((1 && *ptr <= ' ') || *ptr == 127 || *ptr == '/')

or just:

     if (*ptr <= ' ' || *ptr == 127 || *ptr == '/')

Since we are checking that the user has supplied an 8-bit character
between SP (32) and 255, excluding "/" (47) and DEL (127).  The test
we use is therefore valid regardless of the signedness of the char
type on a particular platform.

If you don't like getting the (nonsense) compiler warning in your
compiles, then patch the version of CUPS you ship with OpenBSD -
you already have stated there are other changes that are OpenBSD
specific anyways...

> If you do not want to use signed char * then you have to rewrite the 
> test. On platforms with unsigned chars the first test (*ptr >= 0) is 
> always true.

So what?  The test is harmless, and a compiler that says "foo is
always true) will probably optimize it away anyways...

> You really should not trade convenience for buggy code....

The code is not buggy, it has been written specifically to work
for both signed and unsigned char types without changes.

-- 
______________________________________________________________________
Michael Sweet, Easy Software Products           mike at easysw dot com
Internet Printing and Publishing Software        http://www.easysw.com