Re: setup will have to wait :[

[Christopher Faylor <cgf at redhat dot com>]

On Mon, Mar 26, 2001 at 07:46:35AM +1000, Robert Collins wrote:
>----- Original Message -----
>From: "Egor Duda" <deo@logos-m.ru>
>Subject: Re: setup will have to wait :[
>
>
>> Hi!
>>
>> Saturday, 24 March, 2001 Robert Collins robert.collins@itdomain.com.au
>wrote:
>>
>> RC> 1. Do "non-unix" apps understand the concept of a FIFO file?
>>
>> directly -- not, but they can possibly inherit, say, stdout handle
>> from parent cygwin process, and it can possibly be a handle of fifo.
>> For example, you run
>> "sh -c ./some-script.sh >fifo-to-script-log-processor"
>> and 'some-script.sh' contains non-cygwin apps
>
>Right, so _how do we pass the fd_ to the child process? I'll look into
>spawn and fork. Worst case though: fifos are cygwin only. Best case I
>find some way of tricking the OS into loading up the cygwin code and
>writing "through it". Hmm I wonder if an IFS could do this (How much
>work did I think this was going to be ?!).

They're inherited automatically.  The fd table is copied automatically.
Non-cygwin processes inherit the handles but they won't know what to do
with them since they don't understand cygwin fds.

>> RC> I know that cygwin already has named pipe support
>>
>> ? what do you mean by this? do you mean native calls such as
>> CreateNamedPipe () etc. or something else?
>
>>From comments about sshd and cvs on 9x and the syncronisation problems -
>so I may be off base. I haven't checked the code. Implementing this via
>Named Pipes will be just as tricky as far as non-cygwin programs though
>won't it?.

You can't use named pipes.  You can't create them on Windows 95.

Cygwin doesn't implement named pipes.  Isn't a fifo == a named pipe
anyway?

cgf