ConfigurationError: Misconfigured setting 'package_cache_clean_limit': -1 should be instance of <class 'float'>

[ruzette]

rezconfig.py throws when disabling package_cache_clean_limit to -1 as noted in the the docs inside rezconfig file.

This is written in the docs:

# If > 0, spend up to this many seconds cleaning the cache every time the cache
# is updated. This is a way to keep the cache size under control without having
# to periodically run :option:`rez-pkg-cache --clean`. Set to -1 to disable.
package_cache_clean_limit = 0.5

Setting package_cache_clean_limit = -1 throws:

16:00:38 ERROR    ConfigurationError: Misconfigured setting 'package_cache_clean_limit': -1 should be instance of <class 'float'>

Environment

• OS Rocky Linux 8.9
• Rez version 2.113
• Rez python version: 3.11

To Reproduce

1. Set package_cache_clean_limit = -1 as described in the doc
2. Run rez-config

Expected behavior
It will not throw the error: 16:00:38 ERROR ConfigurationError: Misconfigured setting 'package_cache_clean_limit': -1 should be instance of <class 'float'> and disable the package_cache_clean_limit

Actual behavior
It throws an error:
16:00:38 ERROR ConfigurationError: Misconfigured setting 'package_cache_clean_limit': -1 should be instance of <class 'float'>

Regression
If applicable, state the last known rez version where the bug did not occur (delete this section if N/A).

[beatreichenbach]

I assume this would be fixed by: #1674

As a workaround you can use:

package_cache_clean_limit = -1.0

[JeanChristopheMorinPerso]

@beatreichenbach #1674 wouldn't help much. It's typed as float. See https://github.com/AcademySoftwareFoundation/rez/blob/main/src%2Frez%2Fconfig.py#L452. We could instead change to type to be "int or float".

[beatreichenbach]

I haven't read how the current schema validation works at all, but pydantic can definitely cast int to float:

from pydantic import BaseModel

class Schema(BaseModel):
    package_cache_clean_limit: float

Schema.model_validate({'package_cache_clean_limit': int(-1)})

Again, sorry if this does not apply. I'm just saying using pydantic as a schema validation tool would probably solve this issue, no?