pip install gbp==0.9.8 fails with latest pip

[asottile]

Environment

• pip version: 20.3.3
• Python version: 3.8.5
• OS: ubuntu 20.04

Description

$ pip install gbp==0.9.8
Collecting gbp==0.9.8
  Downloading gbp-0.9.8.tar.gz (150 kB)
     |████████████████████████████████| 150 kB 1.8 MB/s 
  WARNING: Requested gbp==0.9.8 from https://files.pythonhosted.org/packages/85/56/8aaed9186628743910010eb573f54867a5aefec118f59a5f478de34813ef/gbp-0.9.8.tar.gz#sha256=c1f764336546cac7bc91e3cbe45f7e79b73d441eb6d97c84fbd3ba1c74b2c746, but installing version 0.0
ERROR: Requested gbp==0.9.8 from https://files.pythonhosted.org/packages/85/56/8aaed9186628743910010eb573f54867a5aefec118f59a5f478de34813ef/gbp-0.9.8.tar.gz#sha256=c1f764336546cac7bc91e3cbe45f7e79b73d441eb6d97c84fbd3ba1c74b2c746 has different version in metadata: '0.0'

this has been working for quite some time -- it's part of the deadsnakes nightly promotion

Expected behavior

I'm not sure what changed, but I pinned this package defensively and it's suddenly broken

How to Reproduce

pip install gbp==0.9.8

Output

$ pip install gbp==0.9.8
Collecting gbp==0.9.8
  Downloading gbp-0.9.8.tar.gz (150 kB)
     |████████████████████████████████| 150 kB 1.8 MB/s 
  WARNING: Requested gbp==0.9.8 from https://files.pythonhosted.org/packages/85/56/8aaed9186628743910010eb573f54867a5aefec118f59a5f478de34813ef/gbp-0.9.8.tar.gz#sha256=c1f764336546cac7bc91e3cbe45f7e79b73d441eb6d97c84fbd3ba1c74b2c746, but installing version 0.0
ERROR: Requested gbp==0.9.8 from https://files.pythonhosted.org/packages/85/56/8aaed9186628743910010eb573f54867a5aefec118f59a5f478de34813ef/gbp-0.9.8.tar.gz#sha256=c1f764336546cac7bc91e3cbe45f7e79b73d441eb6d97c84fbd3ba1c74b2c746 has different version in metadata: '0.0'

[uranusjr]

It seems like the project is packaged incorrectly. pip got stricter recently about package consistency. You should contact the maintainers of gbp to fix this.

[asottile]

I'm not entirely satisfied with that answer -- there was no warning in 20.2.x for this particular package so it shouldn't go straight to erroring in 20.3.x

[pradyunsg]

This is a change related to the new resolver: see #9203 for more discussion around this change.

I'll close this to avoid having the discussions be spread across various locations.