morotti

@jaraco I saw you fixed the tests on master. I've rebased and tests are green now for this PR. Can you have another look? There is a diffcov workflow complaining...

I am trying to think of what to do without getting into horrible hacks 🤔 🤔 🤔 The caching is tempting as a simple solution but the invalidation is going...

Hello, all green and all comments reviewed, should be good to merge now

I've replaced 1024*1024 with a constant as requested. two constants actually, because the network chunks and file chunks don't necessary need to be the same size and we don't necessarily...

(pingback) ping back to the bug tickets in requests, they've had the issue opened for nearly 10 years to stop using a small chunk size but it was never fixed...

one quick check on older version of pip on older python version the downloading is nearly 3 times faster =) pip version 21.0.1 on python 3.8 default chunk size 10...

Hello, I've made adjustments. We can now reach ~450 MB/s download speed internally, up from ~60 MB before this PR. * Chunks are set to 256 kB. That's sufficient to...

I see failing tests. rebased on master and squashed all the commits. the next build should pass.

(repushing to trigger builds, there seem to be a flaky test on Windows) Thanks, I've adjusted the progress bar to only render for packages > 512 kiB, up from 40...

There seem to be two tests in master that are flaky on Windows. Added in May. FAILED tests/unit/test_utils_retry.py::test_retry_wait[0.015] - assert (669.765 - 669.75) >= 0.015 FAILED tests/unit/test_utils_retry.py::test_retry_time_limit[0.01-10] - assert 11