-
Notifications
You must be signed in to change notification settings - Fork 54
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Feature: more/better logging for greeter.py
, and more reliability
#369
Feature: more/better logging for greeter.py
, and more reliability
#369
Conversation
@@ -87,6 +87,9 @@ def create_session(self, username: str): | |||
def start_session(self, command: List[str], environment: List[str]): | |||
try: | |||
subprocess.check_call(["plymouth", "quit", "--retain-splash", "--wait"]) | |||
except FileNotFoundError: | |||
logging.debug("Failed to stop Plymouth: could not find executable") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That's not going to work, the greeter is always wrapped with the Plymouth executable.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You told me on Matrix that the error is benign. Did I misunderstand something?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The error is indeed benign, but FileNotFoundError
will never actually trigger here.
@@ -122,11 +125,15 @@ def _recv(self) -> Mapping[str, Any]: | |||
class Context: | |||
def __init__(self, user: str, home: Path): | |||
self.user = user | |||
logging.debug("USER: {}".format(self.user)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is probably unnecessary, you can inspect the environment via /proc or systemd or any other tool.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Unless the script is dying after ~100ms. ;)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
In that case you should probably have a stacktrace?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not if it exited with exit(1)
due to misconfiguration. And even if you do, the stack trace won't always tell you what you need to know in my experience.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Then the problem is that the script is doing sys.exit(1)
without providing adequate information.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
No, because again, the stack trace won't always give you what you need -- unless your plan is to catch every exception in the entire program and then do your conditional dump.
The point of the logging system is usually to be as verbose as you could possibly need, and then quiet it by setting the log level higher than DEBUG/TRACE. Usually on WARN or something. Then provide a way to override the debug level if something goes wrong. Since this is a daemon configured by a Nix expression, maybe have the nix expression control a flag that in turn sets the debug level, if you're worried about cluttering up logs?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not to say that exiting with more information isn't a good idea, but dumping the entire execution context only iff the program exits is usually not done because there's a gradual accumulation of state information you need to dump as programs grow in size and complexity. Although this is a trivial Python script, I still think it would make sense to follow that pattern on account of the fact that Python is not a correctness-oriented language and whatever debugging mechanism you supply needs to account for it failing somewhere along the execution chain in a way you didn't expect.
@@ -137,6 +144,7 @@ def _consume_session(self) -> Optional[str]: | |||
helper = HELPER_PREFIX.joinpath('consume-session') | |||
if helper.exists(): | |||
logging.debug('Using pkexec helper') | |||
logging.debug('Helper path: {}'.format(helper)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This should be static.
@@ -145,6 +153,10 @@ def _consume_session(self) -> Optional[str]: | |||
env={'SHELL': '/bin/sh'} | |||
) | |||
next_session = res.stdout.decode('utf-8').strip() | |||
stderr = res.stderr.decode('utf-8').strip() |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Does consume-session ever log to stderr?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not deliberately, but the commands it run might fail, and those commands do log to STDERR. In addition, it's generally a good idea to future-proof this sort of thing anyway IMO.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Actually, we should probably use stdout=PIPE
instead of capture_output
. That way stderr will automatically end up in the journal.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That's fair enough! I think that would work nicely :)
@@ -166,16 +178,19 @@ def _consume_session(self) -> Optional[str]: | |||
return next_session | |||
|
|||
def _find_sessions(self, sessions: Iterable[str]) -> Optional[Session]: | |||
for data_dir in self.xdg_data_dirs + [ '/usr/share' ]: | |||
for data_dir in self.xdg_data_dirs + [ '/usr/share', '/run/current-system/sw/share' ]: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
/run/current-system/sw/share should generally always be on XDG_DATA_DIRS, and if it isn't, we should figure out why.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hmm. It seems in the path by the time I get to the shell or UI, so it's most likely that the environment greeter.py
is running in has something wrong with it. Is the concern here that the path might change?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The concern is that this is unnecessary, and in a year we won't remember why we put it there.
desktop_file = f'{session}.desktop' | ||
wayland_session = data_dir.joinpath('wayland-sessions').joinpath(desktop_file) | ||
x_session = data_dir.joinpath('xsessions').joinpath(desktop_file) | ||
|
||
logging.debug("Checking Wayland session path: {}".format(wayland_session)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'd rather just log on error here, i.e. if session does not exist, log that at WARNING (or even ERROR).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The Wayland session path that it found is useful information for debugging, tho. It's generally a good idea to log I/O paths as TRACE or DEBUG IMO -- as the moment your program is touching other programs/file structures, making the interface points clear can make a lot of confusing errors more obvious.
If you're worried about cluttering up the log files, maybe it shouldn't default at log level DEBuG?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't think it's particularly useful in 99.9% of cases.
IDK, I think it's pretty clear @K900 is not very interested in improved logging, and I'm running out of time that I could be spending on my actual work, so I'mma just peace out and maintain my own fork. Best of luck, peeps. |
@@ -122,11 +125,15 @@ def _recv(self) -> Mapping[str, Any]: | |||
class Context: | |||
def __init__(self, user: str, home: Path): | |||
self.user = user | |||
logging.debug("USER: {}".format(self.user)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
when logging variables in python, best practice is to use %
interpolation.
logging.debug("USER: {}".format(self.user)) | |
logging.debug("USER: %s", self.user) |
see: https://docs.python.org/3/howto/logging.html#logging-variable-data
Changes:
More logs (in the interest of debugging misconfigurations & other issues -- it prints environment variable context, specific paths being checked during early business logic, etc)
Explicitly checks under
/run/current-system/sw/share/
forxsessions
andwayland-sessions
.Replace the scary-looking tracelog dump for
plymouth
executable being missing with a specific, concise error message that explicitly indicates it is not a serious failure mode. (I saw some confusion surrounding this issue.)Closes #368