On Mon, 2002-09-16 at 20:54, Guido Serassio wrote:
> Hi Robert,
>
> I'm working on Your MinGW/Squid patch, and I have some questions:
>
> - What is the correct compile procedure with Cygwin/MinGW ?
I thought I posted that in my email.
I use
CC="gcc -mno-cygwin" srcpath/configure --enable-win32specific (or
whatever the win32 flag is :}) and any other options you usually use.
> - On a previous mail you wrote about _free_osfhnd:
> > > The NT Native port of squid, build with Visual C uses this when linking
> > > to MSVCRT. As a reference in 1999 XEmacs started using this call, but
> > > they defined it locally rather than patching mingw from what I can tell.
> Do you have some reference about, link , etc. ?
I searched google. "Xemacs _free_osfhnd" brings it up.
> - Why do You have moved getopt() from win32lib.c, the new it's a better
> source or is only a organizational move ?
Uh, I forget :]. I grabbed the cygwin getopt.c and getopt.h. I think it
should be separate though, as it's not a win32-specific thing, just a
broken libc thing.
> A final consideration: looking at MinGW sources, it seems to me that the FD
> handling with _free_osfhnd can't work using the MSVCRT.DLL, because the
> function it's not exported from the DLL version of MSVCRT. So, I hope to
> find some workaroud.
Hmm. Well one step at a time :}.
How much reorganisation do you think you would need to use WSASend etc
on network FD's. If we provided calls fdIsSocket(int) and fdIsDisk(int)
and fdIsPipe(int), would that help? I.e. lets separate disk and socket
IO now.
Rob
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:16:33 MST