It seems like Caruso merged this Windows specific test with the Unix counterpart for master and you are right with the language specific error message, perhaps we should use %s in place there as the error error messages seems to come from the function itself.

As for the error itself, that I did not have enough time to dig into here, but does the master one pass?

Yes, master and 7.3 are apparently okay.  Regarding the localized
message: it actually doesn't make sense to raise E_WARNING for
something that succeeded.  Anyhow, now that I had a closer look at
the test[1], I don't think it makes much sense at all.  Looks like
the test tries to increase code coverage, without testing the
actually relevant stuff (for instance, that after enabling crypto
the stream still “works”).

[1] <https://github.com/php/php-src/blob/php-7.3.0beta3/ext/standard/tests/streams/stream_socket_enable_crypto.phpt>

Yeah I was a little curious on that when I was reading over the code, it does indeed seem like it was just for coverage. At least with our more recent point of view in regards to these, we should probably remove the test by itself.

As for testing actual relevant stuff, I guess what we could do for streams stuff, is to maybe make a forked version of already existing tests to enable crypto stuff where it makes sense unless someone steps up to specifically write such.

Yeah, this test has been a headache for some time already. And it's a bit controversial, as creating a socket server and then enabling crypto as a client seems quite odd. Such tests make sense for testing abnormal situation, but in this case it requires too much maintanance. I'll check to backport these changes from 7.3, otherwise this test should be removed.

Thanks.

Thank you Anatol!