new git-annex regression/change in beh? AnnexRepo.info assertion assert(j.pop('file') == f) fails

[yarikoptic]

From daily run on maint -- 18 failures in
https://github.com/datalad/datalad-extensions/runs/603806897?check_suite_focus=true
due to similar (I think all for the same reason) in invocation of git annex info:

  File "/opt/hostedtoolcache/Python/3.7.6/x64/lib/python3.7/site-packages/datalad/support/annexrepo.py", line 1742, in is_under_annex
    files, allow_quick, batch)
  File "/opt/hostedtoolcache/Python/3.7.6/x64/lib/python3.7/site-packages/datalad/support/annexrepo.py", line 1655, in _check_files
    annex_res = fn(files, normalize_paths=False, batch=batch)
  File "/opt/hostedtoolcache/Python/3.7.6/x64/lib/python3.7/site-packages/datalad/support/annexrepo.py", line 1727, in check
    fast=True, **kwargs)
  File "/opt/hostedtoolcache/Python/3.7.6/x64/lib/python3.7/site-packages/datalad/support/gitrepo.py", line 315, in newfunc
    result = func(self, files_new, *args, **kwargs)
  File "/opt/hostedtoolcache/Python/3.7.6/x64/lib/python3.7/site-packages/datalad/support/annexrepo.py", line 2432, in info
    assert(j.pop('file') == f)
AssertionError

annex 8.20200330+git119-gcee6b344b-1~ndall+1
attn: @joeyh

[yarikoptic]

underlying issue -- git-annex started to "normalize" paths a bit:

(Pdb) p j
{'command': 'info test-annex.dat', 'size': '28', 'success': True, 'key': 'SHA256E-s28--2795fb26981c5a687b9bf44930cc220029223f472cea0f0b17274f4473181e7b.dat', 'error-messages': [], 'present': False, 'file': 'test-annex.dat'}
(Pdb) p f
'./test-annex.dat'
(Pdb) p files
['./test-annex.dat', './not-committed.txt', './INFO.txt']

so it went from provided './test-annex.dat' to 'test-annex.dat' .

[yarikoptic]

but not always:

(Pdb) p j['file']
'./.datalad/somefile'
(Pdb) p f
'./.datalad/somefile'

[yarikoptic]

We have a workaround and I have a vague memory of reporting something like that against git-annex, but we do not even have a test to verify correct operation if we RF things up... but not sure what else to keep this issue open for, so will close to beat @mih in the number of closed issues per day