Fix --print-exception-stacktrace not invalidating pantsd

[Eric-Arellano]

Before, if a command failed and you run it again with --print-exception-stacktrace, it would make no difference because the daemon's cache was being used.

Now, we force pantsd to restart so that the option works properly.

[ci skip-rust-tests]

[coveralls]

Coverage remained the same at 0.0% when pulling 7616485 on Eric-Arellano:fix-pantsd into 9ca7b52 on pantsbuild:master.

[stuhood]

See #10035 for the explanation of when fingerprint=True vs daemon=True is relevant, but: afaict exceptions before the scheduler is created should always result in a full stacktrace, so it shouldn't be necessary to restart pantsd (ie, daemon=True) for this option.

On the other hand, having this marked fingerprint=False means that we won't recreate the Scheduler if this value changes... and that is a bug, because the Scheduler does consume this option. So I think that you probably just want to mark this fingerprint=True (which should be equivalent to removing the argument).

[Eric-Arellano]

Sounds good. This morning I was having issues with fingerprint=True actually causing a change in behavior - it did change the behavior, but it looked like portions of the stacktrace were still being printed (less than normal). I didn't capture that output, sorry.

I can't reproduce it though, so looks like this is working correctly now.

[Eric-Arellano]

I can't reproduce it though

That would be because I changed the wrong option. Oops.

This is the output I get when fingerprint=True, daemon=False. This was after running with --print-exception-stacktrace already.

▶ ./pants --no-pantsd --no-print-exception-stacktrace lint src/python/pants/util/strutil.py
17:18:34 [ERROR] 1 Exception encountered:

17:18:41 [ERROR] 1 Exception encountered:

  ValueError:
Traceback (most recent call last):
  File "/Users/eric/DocsLocal/code/projects/pants/src/python/pants/bin/local_pants_runner.py", line 288, in run
    engine_result = self._run_v2()
  File "/Users/eric/DocsLocal/code/projects/pants/src/python/pants/bin/local_pants_runner.py", line 176, in _run_v2
    return self._maybe_run_v2_body(goals, poll=False)
  File "/Users/eric/DocsLocal/code/projects/pants/src/python/pants/bin/local_pants_runner.py", line 199, in _maybe_run_v2_body
    poll_delay=(0.1 if poll else None),
  File "/Users/eric/DocsLocal/code/projects/pants/src/python/pants/init/engine_initializer.py", line 135, in run_goal_rules
    goal_product, params, poll=poll, poll_delay=poll_delay
  File "/Users/eric/DocsLocal/code/projects/pants/src/python/pants/engine/internals/scheduler.py", line 558, in run_goal_rule
    self._raise_on_error([t for _, t in throws])
  File "/Users/eric/DocsLocal/code/projects/pants/src/python/pants/engine/internals/scheduler.py", line 528, in _raise_on_error
    wrapped_exceptions=tuple(t.exc for t in throws),
pants.engine.internals.scheduler.ExecutionError: 1 Exception encountered:

  ValueError:

If I have daemon=True, I get the expected result:

▶ ./pants --no-pantsd --no-print-exception-stacktrace lint src/python/pants/util/strutil.py
17:18:34 [ERROR] 1 Exception encountered:

  ValueError:

(Use --print-exception-stacktrace to see more error details.)

[stuhood]

Hm. It looks like that's because this is a sneaky global-mutable option on the ExceptionSink:

pants/src/python/pants/init/logging.py

Lines 70 to 76 in 9ca7b52

	class ExceptionFormatter(Formatter):
	"""Uses the `--print-exception-stacktrace` option to decide whether to render stacktraces."""
	def formatException(self, exc_info):
	if ExceptionSink.should_print_exception_stacktrace:
	return super().formatException(exc_info)
	return "\n(Use --print-exception-stacktrace to see more error details.)"

... (which I added. *sigh) would you mind adding a comment/TODO explaining that and setting daemon=True for now?

[stuhood]

With the comment, please. Thanks, and sorry for the trouble!