AW: [Zlib-devel] CE Port

[Cosmin Truta]

On Wed, 18 Jun 2003, Alexander Mai wrote:

> > I looked a little more at the Windows CE port, and I would like to
> > bring in some objections and suggestions.
>
> You are talking about the existing, full CE port on that
> given URL, right?!

I am talking about
  http://www.tenik.co.jp/~adachi/wince/zlibce/index.html

> My recent proposal is not to use it, but to incorporate
> the set of patches being provided by me. (a first set was
> already sent to this list, more to come when I have a bit more
> time...)

Now I realize that you provided a different patch, to a newer version of
Windows CE, that has the standard C library. This is excellent!

I am sorry that my previous message on this topic mislead some people.


> > As I see, Windows CE doesn't support a minimum of decent portability.
> > ***What's good in using C, if it isn't possible to use the standard C
> > library properly?***
>
> A somehow funny statement. What would you use on a system where
> only a non-ANSI C implementation exists??

If one has an alternative that is technically better than C (that is,
a familiar programming language that offers the same performance as C,
but a higher level of expression), and the portability reason does not
exist, then there is not much reason left in the favor of C. But if C
is your only choice, then, you are right, you are stuck with it.
My point was that a C platform without the standard library should be
treated like a non-C platform - in the same manner as the non-C ports,
found in the contrib/ directory.


> > - If there is no sprintf() or snprintf(), there should be wsprintf()
> > on Windows CE. (Is this correct?)
>
> The "C API" exists for char and the unicode versions (at least
> as far as available at all...).
> The system API however is pure UNICODE.

I was talking about the other port for the old CE. I am glad to know
that we don't have to worry about non-standard library workarounds
anymore.


I wish to make one suggestion, though: I see that rewind() is
unavaliable on CE. The way you provide it via a macro will break as soon
as it will become available on a later version of CE. I suggest to
replace rewind(s->file) with fseek(s->file, 0L, SEEK_SET) directly in
gzrewind().

--- zlib-1.2.0.1/gzio.c		Thu Apr 03 00:27:56 2003
+++ zlib-1.2.0.1-CE/gzio.c	Wed Jun 18 15:30:00 2003
@@ -811,7 +811,7 @@
     s->crc = crc32(0L, Z_NULL, 0);

     if (s->startpos == 0) { /* not a compressed file */
-        rewind(s->file);
+        fseek(s->file, 0L, SEEK_SET); /* rewind() is not always available */
         return 0;
     }

The difference between rewind(fp) and fseek(fp, 0L, SEEK_SET) is that
rewind() clears the error and the EOF indicator, while fseek() only
clears the EOF indicator. This is not relevant to zlib's behavior.


Best regards,
Cosmin