Provide current log line context when a parsing exception is encountered

[stennie]

Unexpected log parsing exceptions currently die with a stack trace. It would be helpful to catch the exception and report the offending log line and any other available context.

[virajut]

Hi @stennie,

Can you share more details about this issue?

[stennie]

@virajut See #764 for discussion that led to the creation of this suggestion.

Unexpected exceptions can currently result in a stack trace which would benefit from additional context (like the current log line being parsed) to help understand and resolve issues faster. The specific example was problematic JSON that has now been fixed, but I've encountered this with other logs (usually accompanying bug reports on GitHub 😄 ). Without knowing the offending log line, there's a bit of guesswork involved based on the stack trace and filtering of the original log file.

Exceptions probably need to be caught via the log iteration in util/logfile.py which is used by all of the log tools.

I don't have any examples handy that will trigger an exception at the moment (you'd have to find or construct some for tests). I was expecting there might be new bug reports that would motivate creating & testing this change, but there haven't been any so far.

Regards,
Stennie