running `urllib.request.urlopen` multiple times causes the memory usage to increase with each run, even after calling `close()` on the request or using `del` on the result

To recreate this problem, run code:

    import urllib.request
    def ip():
        r = urllib.request.urlopen('https://api.ipify.org')
        b = r.read()
        r.close()
        print(len(b))
        return b

    for i in range(1000):
        result = ip()
        del result

Even though `len(b)` has a maximum value of 15 (for this url at least), the memory increases with each run by 200KB - 1MB

Just to note: I'm using Python 3.7.4 on Windows 10 (64bit), the issue exists on both the 32bit/64bit python versions I could test on.

JFYI, Not reproducible on macOS.

Is this related to https://bugs.python.org/issue35941 given that it's related to windows. Similar report https://bugs.python.org/issue37498

Yes, https://bugs.python.org/issue37498 is indeed the same issue

Closing as a duplicate of issue35941

My Windows 10 (64bit). There is no such problem in 3.6.1, but 3.7.0 has. And when requesting `http://...`, both of them are normal.

By the way, the related issues are all closed, is it supposed to reopen one?

I can confirm that testing with http instead of https this problem doesn't exist, on 3.7.4.

Are the issues closed because this problem is fixed in the upcoming 3.7.5, or 3.8.0?

Yes, the problem will be fixed by 3.7.5.

In the mean time you can work around the issue by creating a single SSLContext object with ssl.create_default_context() and reusing it for all requests. It's also much more efficient to reuse a single context for all requests.

I appreciate the suggestion and the update, thank you!

resolution: duplicate
stage: resolved