Mailing-List: contact dev-help@apr.apache.org; run by ezmlm
Precedence: bulk
Message-ID: <001801c1a831$9f1a2840$7f89bfd5@armada>
From: "Mladen Turk" <mturk@mappingsoft.com>
To: "William A. Rowe, Jr." <wrowe@rowe-clan.net>,
	<dev@apr.apache.org>
References: <000701c1a7fb$37048ec0$5f00000a@armada>
 <026b01c1a81b$6ce0cde0$95c0b0d0@v505>
Subject: Re: [PATCH] Support WINNT builds and initial WINCE support
Date: Mon, 28 Jan 2002 20:25:52 +0100
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit


----- Original Message -----
From: "William A. Rowe, Jr." <wrowe@rowe-clan.net>
To: <mturk@mappingsoft.com>; <dev@apr.apache.org>
Sent: Monday, January 28, 2002 5:44 PM

The reason for changing the os enumeration was to be able to add the extra
LSB bit that indicate if the OS has UNICODE support (see the
APR_OS_IS_UNICODE macro), but I think you figured that out.

> +#define APR_HAS_WINSOCK2    1
>
> You use this below, but I'm unclear.  Where are you 'toggling' this?
Where are
> we now supporting non-winsock2?  If CE is a single socket API - is it
effectively
> winsock or winsock2?  And what are the rest of the compatibility patches
we have
> to put it to support the old winsock?
>
> I'm very uncomfortable in this territory, there we a number of good
reasons for
> dropping winsock support in Apache, and this could be a huge step
backwards.
>
> In any case, APR_HAS_WINSOCK2 should be a private internal detail, since
it's
> another symbol that must land in apr.hw/apr.h.in.  If we can't do so, it
seems
> like you were reaching for APR_HAVE_WINSOCK2_H as a symbol name.

The CE lacks the Winsock2  (at least 3.0) so the only way was to use the old
winsock
So, If some further versions introduce the Winsock2, we could easily drop
that out.
As far the symbol name the APR_HAVE_WINSOCK2_H is fine

>  #define APR_HAS_USER              1
>  #define APR_HAS_LARGE_FILES       1
>  #define APR_HAS_XTHREAD_FILES     1
> +#define APR_HAS_PIPES             1
> +#define APR_HAS_STDIO             1
> +
>
>
> I'm confused, didn't see these symbols used in your patch.  Again, a
change
> that must happen to both apr.hw and apr.h.in.  Did you mean STDIO to be
> APR_HAVE_STDIO_H?  And we don't want to proliferate symbols that aren't
tested,
> if you will need this 'soonish', wait until you have a patch that
demonstrates
> the need for the new symbols.

The APR_HAVE_STDIO is used for CE to switch out all the console code, and
APR_HAS_PIPES to switch out the pipes code.
Now, since CE lacks the pipes, I was thinking to try to implement that using
shmem, but that can be dropped

Next I'll try to sweep again throug the sources and #ifdef all the
nosuported headers, and perhaps delete the one that are allready included
either from apr.h or misc.h

MT.