Run curl_setopt_ssl.phpt on Windows

[cmb69]

There seems to be no particular reason to skip this test on Windows.

While we're at it, we also remove the check for proc_open() since this function is required (and checked) by run-tests.php anyway.

[Ayesh]

I checked on Windows (openssl 3.3 custom build, Curl 8.10.1), and the test indeed passed on Windows. I'd say lets do it and enable this for Windows.

I noted that ext/openssl/tests/stream_server_reneg_limit.phpt also skips on Windows, with 1552d6ae7bb from 10 years ago mentioning openssl binary availability being unreliable.

[cmb69]

@Ayesh, indeed the test passes for me locally, but might hang in CI, possibly because the pipes are never read, but openssl produces output. Using socket instead of pipe should fix that.

However, that might also be the problem on FreeBSD:

php-src/ext/curl/tests/curl_setopt_ssl.phpt

Line 12 in 87e7915

if (PHP_OS === 'FreeBSD') die('skip proc_open seems to be stuck on FreeBSD');

So maybe we should just read stdout and stderr.

I'll have a look at ext/openssl/tests/stream_server_reneg_limit.phpt shortly. Thanks for mentioning!

[cmb69]

Hmm, there's still something wrong. Needs further investigation.

[cmb69]

Rebased and force-pushed to trigger CI again (last failure looks actually completely unrelated).

[cmb69]

At least one of these tests stalls in CI; could reproduce locally too, but have not been able to resolve. More important things to do right now, so putting this to backlog.

[cmb69]

Gosh! When running the test (locally) the first time, it stalls. Afterwards, the openssl s_client process is still running, so running the test again works. proc_open() would not fail, because CreateProcessW doesn't wait until the process has finished its initialization, but returns success early. Why the test hangs for the first run, and why openssl is not closed, see 430a04f.

[cmb69]

Seems to work now. Test failure is unrelated; I've left a comment at #16875 (comment).