Re: [Sbcl-devel] SBCL on Cygwin: is there a special build configuration ?

[Ilya Z. <mas...@gm...>]

Nikodemus Siivola wrote:

First of all, thank you for reply.
> What exactly is it that CLORB needs? Does it call unix-select
> directly? A pointer to CLORB source where the dependency occurs would
> be appreciated.
>   

Yes, it does call `unix-select' directly. This seems to be its
ideology not to use any adapter libraries like USOCKET etc.

This happens (AFAIR this is the only place as CLORB creates
its own wrappers) in `clorb-sysdep.lisp' somewhere around line
# 570, the code is like :

(defmacro %unix-select ....

CLORB home page is  http://clorb.sourceforge.net

CVS repo is

cvs -d:pserver:ano...@cl...:/cvsroot/clorb login
cvs -z3 -d:pserver:ano...@cl...:/cvsroot/clorb co clorb


> Please post your test-case. (So we can agree that it tests the right
> thing, and hopefully so that someone can use it to drive the Win32
> work needed.)
>
>   
I'll attach. But please note that the test is NOT automatic,
the idea is to wait for availability of input and read. If
select works, the select call is long, but read is short,
if select does not, the opposite happens.

There is also a simple echo socket server written in Python,
included in the archive, it is used in the test.
> Providing supported a select() interface for Win32 as well as *nix
> platforms is certainly possible, but it is not going to be
> UNIX-SELECT. (Maybe SB-SYS:FD-STREAM-SELECT?)
>   
As I figure out, cygwin provides same functionality as
`select' on any unix. I hoped to get all working by recompiling
for cygwin. Providing the full unix-like APIs in win32 is another big
story, I think.

Anyway, thanks.