Re: PATCH: LWPng-alpha on Windows NT

Gurusamy Sarathy (gsar@engin.umich.edu)
Fri, 03 Jul 1998 19:07:31 -0400 

On Fri, 03 Jul 1998 12:27:19 PDT, Blair Zajac wrote:
>> I have some touble with this.  I would really like to avoid adding XS
>> code to LWP and I don't really like the LWP::Extras name either.
>> Wouldn't it be much better to just say that you need IO-1.19 installed
>> to make it work on WinNT?
>
>What about older versions of Perl?  Are we going to require an upgrade in
>either IO or Perl to use LWPng-alpha?  In my experience, putting a newer
>IO in an older Perl hasn't always worked too well.

I don't see much problem with asking people to use LWPng with the latest
version of perl available for win32.

>> 
>> Does IO-1.19 work on WinNT?

I haven't tried (one of the things on my evergrowing todo lists).

>> ... or just test $^O and hardcode some EINPROGRESS that works for
>> WinNT?
>
>We could do that.  EINPROGRESS on WinNT is 10036.  I don't know for sure
>if the socket calls will return EINPROGRESS if the socket call will block.
>Somebody else may need to answer this.

I think EINPROGRESS will be returned only if a blocking call is already
being processed for the process.

>The latest Errno in perl 5.004_69 package doesn't seem to find EINPROGRESS.
>I don't have an NT box here at home, but that's what I recall from getting
>LWPng-alpha going on NT at work.  I'm by no means a WinNT expert here, but
>here are my comments.  Errno on WinNT did not appear to set EINPROGRESS.
>In examining the winsock.h, it looks somewhat like this
>
>#define WSABASEERROR 10000
>.
>.
>.
>.
>#define WSAEINPROGRESS WSABASEERROR + 36
>.
>.
>.
>/* We (Microsoft) decided not to define these for some reason. */
>#if 0
>.
>.
>#define EINPROGRESS WSAEINPROGRESS
>.
>.
>#endif
>
>The whole file looks like this and defines a whole series of WSA constants.
>We may need to look for the WSA constants and convert these to the standard
>E constants.  Do we use these WSA constants?

The current Errno build doesn't even look for winsock.h for the socket
Econstants.  I think this should be done.

 - Sarathy.
   gsar@engin.umich.edu