fix requests auto decode file

[codeskyblue]

only happens when Content-Encoding: gzip

Pull Request Check List

Resolves: #4523

[dimbleby]

poetry downloads probably millions of tar.gz files every day without issue: the real problem here is that your server is misconfigured.

the proposed fix will break things for people whose servers are correctly configured and who are (admittedly unnecessarily) zipping the tar.gz for download. pip tried the same fix years ago and saw exactly that - pypa/pip#1435

Having said that, they do now seem to have landed in a place where they try to compensate for misconfigured servers by using Accept-Encoding: Identity - see all of the code in https://github.com/pypa/pip/blob/main/src/pip/_internal/network/utils.py

personally I would prefer that you corrected your server configuration and we left poetry alone, but if you want to pursue this path here then there is more to do.

edit: on reflection I think that just setting Accept-Encoding: Identity - and making no other change - should be reasonable. whl and tar.gz files are already compressed so telling servers not to compress them again during transfer is just saving compute on both sides

[codeskyblue]

poetry downloads probably millions of tar.gz files every day without issue: the real problem here is that your server is misconfigured.

the proposed fix will break things for people whose servers are correctly configured and who are (admittedly unnecessarily) zipping the tar.gz for download. pip tried the same fix years ago and saw exactly that - pypa/pip#1435

Having said that, they do now seem to have landed in a place where they try to compensate for misconfigured servers by using Accept-Encoding: Identity - see all of the code in https://github.com/pypa/pip/blob/main/src/pip/_internal/network/utils.py

personally I would prefer that you corrected your server configuration and we left poetry alone, but if you want to pursue this path here then there is more to do.

edit: on reflection I think that just setting Accept-Encoding: Identity - and making no other change - should be reasonable. whl and tar.gz files are already compressed so telling servers not to compress them again during transfer is just saving compute on both sides

updated code according to your request.

[dimbleby]

looks ok to me (once you fix up the linting).

Does it work in getting your server to behave itself?

[codeskyblue]

looks ok to me (once you fix up the linting).

Does it work in getting your server to behave itself?

I fix this issue by remove Content-Encoding from my server (flask based pypi mirror which is not support Accept-Encoding: Identity)

[codeskyblue]

how to fix the linting? any tools?

[dimbleby]

how to fix the linting?

Per the errors! ruff and black

[codeskyblue]

To make the pr simple, I removed the tests code

[Secrus]

To make the pr simple, I removed the tests code

No chance for a PR to be merged without tests of any kind. If you don't want to write tests for that, close this PR.

[radoering]

To make the pr simple, I removed the tests code

That just sounds so wrong. 😄

Nevertheless, I think the change itself makes sense, so I just converted your test from pytest-httpserver to httpretty, which was quite simple.

[radoering]

pre-commit.ci autofix

[github-actions]

This pull request has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.