On Wed, 25 Feb 2004, David Byron wrote:

> Here's a patch to remove winsock 2 dependence. Some of the
> changes here are duplicates of what I sent earlier to get
> MSVC 6 to build. It's probably best to deal with that first
> and then deal with this, but I'd love to get this change in
> so I wanted to give everyone a chance to see it as early as
> possible.

I've attached a new patch now that Daniel has committed the other MSVC
changes (thank you).

> I don't really have a good setup for testing telnet. I'm
> working on it. I'll also run the testsuite soon.

I've done some very basic telnet testing. Everything seems fine for me.
I ran the testsuite and came up with the following:

srcdir=. /usr/bin/perl -I. ./runtests.pl -a
********* System characteristics ********
* curl 7.11.0-CVS (win32)
* libcurl/7.11.0-CVS OpenSSL/0.9.7a
* Host: DByron
* System: CYGWIN_NT-5.1 DByron 1.5.7(0.109/3/2) 2004-01-30 19:32 i686
unknown unknown Cygwin
* Server SSL: OFF
* libcurl SSL: ON
* libcurl debug: OFF
* valgrind: OFF
*****************************************
TESTDONE: 143 tests out of 155 reported OK: 92%
TESTFAIL: These test cases failed: 20 63 500 501 502 503 504 505 506 507 508
510
TESTDONE: 172 tests were considered.
TESTINFO: 17 tests were skipped due to these restraints:
TESTINFO: "no HTTPS server" 8 times (300, 301, 302, 303, 304, 305, 306, 509)
TESTINFO: "no FTPS server" 4 times (400, 401, 402, 403)
TESTINFO: "curl lacks netrc_debug support" 5 times (130, 131, 132, 133, 134)

All the 500-series tests failed because the test programs don't exist --
there's no MSVC Makefile for them. One day maybe :)

Test 63 failed for me before. The reason is that the win32 curl reads
environment variables in a different way than runtests.pl sets them. I
haven't figured out a good way around this.

Test 20 is a new failure for me. I haven't done the full cycle of backing
out my change and re-running the tests to confirm, but...The details are:
test 020...[attempt connect to non-existing host name]
curl returned 28, 6 was expected
 error FAILED

log/stderr20 contains:
* Resolving host timed out: non-existing-and-weeeeeeirdname
* Closing connection #0
curl: (28) Resolving host timed out: non-existing-and-weeeeeeirdname

I seem to remember some discussion of timeouts, but haven't been following
them closely. Is the solution to just change tests/data/test20 to expect
error code 28?

Thanks much for taking a(nother) look at this patch.

-DB