tests: fix win32

[trim21]

Summary

There are several test cases broken on windows.

Checklist

• I understand that this PR may be closed in case there was no previous discussion. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change.
• I've updated the documentation accordingly.

[trim21]

coverage failed because this:

httpx/httpx/_multipart.py

Lines 127 to 134 in fa6dac8

	if isinstance(fileobj, io.StringIO):
	raise TypeError(
	"Multipart file uploads require 'io.BytesIO', not 'io.StringIO'."
	)
	if isinstance(fileobj, io.TextIOBase):
	raise TypeError(
	"Multipart file uploads must be opened in binary mode, not text mode."
	)

io.StringIO is a subclass of io.TextIOBase

[trim21]

I just use pytest's tmp_path to create a real file on disk. pytest will clean it up after test.

[tomchristie]

We don't need this skipif now, right?

[trim21]

we still need this skipif because path seprator / is double escaped on this tests, don't know why....

[tomchristie]

Can you remove the skipif, so we can see the exact error in CI?

[trim21]

E
E         At index 1 diff: ('httpx', 10, "load_verify_locations cafile='C:\\\\Users\\\\Trim21\\\\proj\\\\httpx\\\\.venv\\\\Lib\\\\site-packages\\\\certifi\\\\cacert.pem'") != ('httpx', 10, "load_verify_locations cafile='C:\\Users\\Trim21\\proj\\httpx\\.venv\\Lib\\site-packages\\certifi\\cacert.pem'")
E
E         Full diff:
E           [
E               (
E                   'httpx',
E                   10,
E                   'load_ssl_context verify=True cert=None trust_env=True http2=False',
E               ),
E               (
E                   'httpx',
E                   10,
E                   'load_verify_locations '
E         -         "cafile='C:\\Users\\Trim21\\proj\\httpx\\.venv\\Lib\\site-packages\\certifi\\cacert.pem'",
E         +         "cafile='C:\\\\Users\\\\Trim21\\\\proj\\\\httpx\\\\.venv\\\\Lib\\\\site-packages\\\\certifi\\\\cacert.pem'",
E         ?                      ++     ++          ++    ++         ++       ++   ++                 ++       ++
E               ),
E           ]

[tomchristie]

I assume we could drop this, and this particular test would actually pass okay on windows?

(Also, I think we might want to raise a related refactoring issue to remove these /slow_response tests completely.)

[trim21]

this test may pass on windows.

[trim21]

it's a time accuracy problem on windows. we could increase sleep time in /slow_response to several seconds to fix this.

[trim21]

I'm wrong about this, looks like increase sleep time in /slow_response won't help testing write timeout

[tomchristie]

We probably don't need this conditional once we've fixed up all the tests to pass okay on windows, right?

[trim21]

scritps/coverage only works on unix

[tomchristie]

I'm not sure - I'm pleased to see us fix the test cases up for supporting windows, tho do we want to double the time our test runs take to complete?