719861 - Talos should have an option to dump configuration

Status: RESOLVED FIXED

(Testing :: Talos, defect)

Description

[Jeff Hammel]

bug 704654 will lend the ability to run talos in one step vs the
current two step process:

1. generate .yml configuration
2. run_tests.py $CONFIGURATION.yml

However, we want the ability to have and use configuration files.  To
that end, a --dump flag should be added that will dump the
configuration to a specified file.

Configuration should go in its file to be consumed by run_test.py

Comment 1

[Chris Manchester (limited bugmail, email directly)]

Attachment: Adds --dump option to run_tests.py using config writer from PerfConfigurator

Adds --dump flag for run_tests.py that uses config writer from PerfConfigurator.

This will write config options provided to run_tests.py, but is unaware of any config that might already be in another config file. This could be a strange behavior as long as we are using a two-step process, however this brings us very close to being able to run tests without a separate step for PerfConfigurator.

Comment 2

[Joel Maher ( :jmaher ) (UTC -8)]

Comment on attachment 591701 [details] [diff] [review]
Adds --dump option to run_tests.py using config writer from PerfConfigurator

Review of attachment 591701 [details] [diff] [review]:
-----------------------------------------------------------------

the only reason for the f- is the need to handle the remotePerfConfigurator case.  Otherwise this looks great.

::: talos/run_tests.py
@@ +421,5 @@
> +  
> +def dump_config(**options):
> +    """uses PerfConfigurator to write preferences to file specified to --dump flag"""
> +    config = PerfConfigurator.PerfConfigurator(**options)
> +    config.writeConfigFile()

do we need a function for this?  What about the remotePerfConfigurator stuff?

@@ +624,5 @@
>      utils.setnoisy(1)
>  
> +  # Write config to a file specified to --dump flag
> +  if options.dump_file:
> +    options.outputName = options.dump_file

any thoughts on just using --output and avoiding a --dump argument?  --dump provides clarity, but seems redundant

Comment 3

[Chris Manchester (limited bugmail, email directly)]

Thanks for the feedback Joel. I'm pretty close to what you've recommended here, but I think this will require what is specified here https://bugzilla.mozilla.org/show_bug.cgi?id=720773 in order to parse for both remotePerfConfigurator and PerfConfigurator in a quasi-intelligent way. I'll take a look at that bug first.

Comment 4

[Jeff Hammel]

Actually I plan on just using the existing -o switch vs --dump so the title of the bug is a bit of a misnomer

Comment 5

[Jeff Hammel]

fixed with bug 704654