docs issue or regression? retries ignored

[ptrebert]

Snakemake version
7.19.1

Describe the bug
The rule-specific retries directive is ignored.

Minimal example

rule foo:
    output:
        "foofile.txt"
    retries: 0
    run:
        print("running")

Running the above with

snakemake -c1 -d wd/ --restart-times 5

will restart the failed rule 5 times.

Additional context
The way I understand the documentation ...

Note that it is also possible to define retries globally (via the --retries command line option). The local definition of the rule thereby overwrites the global definition.

the retries directive in a rule should always take precedence.

This is related to #1649 (and somewhat to #367 and #500)

[pvandyken]

This is a bug. The current implementation is not correctly handling retries: 0. Compare to an identical reprex where retries is set to 1.

Interestingly, such an experiment reveals a second bug, specifically related to the use of the run directive. Snakemake evaluates such rules in a subsnakemake instance, but the retries directive is getting honoured both within the parent and child snakemake instances, leading to 3 retries in total.

I'll create a new issue for the second bug, using the adapted minimal example.