Path is not represented in current version locations

[harpener]

Hi guys.

Describe the bug

Cannot generate migration in project with multiple heads because location paths do not match.

Expected behavior

Migration is generated.

To Reproduce

alembic revision --autogenerate --head "after@head" -m "my_migration"
alembic revision --autogenerate --head "before@head" -m "my_migration"

Error

FAILED: Path C:\...\alembic\after is not represented in current version locations
FAILED: Path C:\...\alembic\before is not represented in current version locations

Versions.

• OS: Windows 11 Pro, 22H2, 22621.1992
• Python: 3.8.17
• Alembic: 1.11.1
• SQLAlchemy: 1.4.48
• Database: PostgreSQL 9.6.2

Additional context

I use two heads - after and before. I have following in alembic.ini:

version_locations = %(here)s/alembic/before %(here)s/alembic/after

I noticed that my Python returns following:

>>> import os.path

>>> os.path.abspath("")
'c:\\...'

>>> os.path.realpath("")
'C:\\...'

As you can see, the letter-case of drive letter is different which is where the problem comes from.

I haven't encountered such error in past, but I recently installed the OS anew and updated Python to its latest security fix, if it helps.

Do you have any idea what is wrong?

Have a nice day!

[CaselIT]

Hi,

What are using as the setting of version_path_separator?

[zzzeek]

it makes no sense that the "c:" is changing case

[zzzeek]

anyway, patches welcome for whatever approach anyone thinks is the "correct" way to do this that passes tests and does not break case sensitive paths on other OSes

[harpener]

It's not the real path, so it can make sense in workspace context. It's the same with symlinks, I am confident that if I put a symlink to a directory in the config file, it will not match the version locations again, even in linux. IMO all that is needed is to translate the %(here)s using the real path, since it is compared to a real path later as well. Thanks for the input.

[CaselIT]

that may be an option. I'll take a look to see if it's a possible change

[zzzeek]

see #1330

[CaselIT]

this seems actually an issue so re-opening it as issue #1366