For

shell_command

, timeouts with

execution_dependencies

that are also

shell_command

entries can vary. Suppose that there are the

shell_command

entries shown below:

shell_command(
    name="a",
    command="./a.sh",
    timeout=120,
)

shell_command(
    name="b",
    command="./b.sh",
    timeout=120,
)

shell_command(
    name="c",
    command="./c.sh",
    execution_dependencies=[
        ":a",
        ":b",
    ],
)

The timeout of

c

assumes the default (30 seconds). Would it make sense for

c

to assume a timeout of 30 seconds given that

a

and

b

have a timeout greater than 30 seconds? Or do you have to account for how long

a

and

b

could take in figuring out what the timeout of

c

should be and incorporate that into

timeout

for

c

manually?

In the absence of an answer from someone knowledgeable you might try commands like

sleep 5

and appropriate adjustments of timeouts to test this and find out.

Isn't c depending on the outputs of a and b? Why would it need to understand their timeouts? By the time c runs, it already has run the commands, no?

I guess the behavior with

execution_dependencies

is what I wasn't certain about. So the timer for the timeout of

c

starts after all

execution_dependencies

(

a

and

b

) finish?

Essentially. It's more like "by the time it goes to run c, by nature of that dependency, a and b would've already ran so that their outputs can be used for C's execution

Yes, the timeouts are for the specific

command

, unrelated to any dependencies

Makes sense, thanks for the clarification