refactor: Move tool name to FlowCfg#149
Conversation
AlexJones0
left a comment
AlexJones0
left a comment
There was a problem hiding this comment.
I agree that this is definitely an easy stop-gap for now, but I checked your strong suspcicion and can confirm that in all the different flows this attribute is defined. SimCfg and OneShotCfg set it directly from args.tool, FormalCfg, LintCfg and SyncCfg inherit it from OneShotCfg, and CdcCfg & RdcCfg inherit it from LintCfg.
So this could be added to FlowCfg.__init__ as self.tool: str | None = args.tool (since the --tool argument can be unspecified), and then this runtime check should instead be that the value is not None. If you don't mind the little extra effort, could you make this change?
rswarbrick
force-pushed
the
tool-field-at-run-time
branch
from
c86c43f to
29986b0
Compare
rswarbrick
changed the title
|
@AlexJones0: You're completely right! Thanks! I think I've tweaked everything to match now? |
| if self.sim_cfg.tool is None: | ||
| msg = ( | ||
| "No tool selected in job configuration. It must either be " | ||
| "specified in the hjson or passed with the --tool argument." |
There was a problem hiding this comment.
Minor nit: extra space
| "specified in the hjson or passed with the --tool argument." | |
| "specified in the hjson or passed with the --tool argument." |
| # At this point, we have finished running a tool, so we know that | ||
| # self.sim_cfg.tool must have been set. | ||
| if self.sim_cfg.tool is None: | ||
| raise AssertionError("sim_cfg.tool cannot be None now.") |
There was a problem hiding this comment.
Minor nit: AssertionError is typically intended for use in the built-in assert statement, I would suggest using a RuntimeError here instead?
There was a problem hiding this comment.
Ah, that makes sense. I'll switch over to that. I think I was intending it to mean "Oh no! Bug in tool" (as opposed to "Naughty user: that's a silly thing to do."), but agree that AssertionError isn't the right way to do that.
Is there a standard way to make this distinction in Python code?
This value was always actually supplied (either directly in SimCfg or through OneShotCfg). Move it to the base class and consume it in just one place. A specific type looks like there might be problems where we try to run a job without having decided on the tool. The second instance of this (in CovReport.post_finish) can definitely never happen, but the example in Deploy.get_job_spec is less obvious to me, so I've put in a runtime check. Signed-off-by: Rupert Swarbrick <rswarbrick@lowrisc.org>
rswarbrick
force-pushed
the
tool-field-at-run-time
branch
from
29986b0 to
52081e8
Compare
2 participants
This value was always actually supplied (either directly in SimCfg or
through OneShotCfg). Move it to the base class and consume it in just
one place.
A specific type looks like there might be problems where we try to run
a job without having decided on the tool. The second instance of
this (in CovReport.post_finish) can definitely never happen, but the
example in Deploy.get_job_spec is less obvious to me, so I've put in a
runtime check.