Alan Jenkins <alan.christopher.jenk...@gmail.com> added the comment:

Oops. I assumed logging.raiseExceptions == True (the default) actually raises 
exceptions, but it doesn't.  It prints the exception to stderr and continues.

E.g. traditional UNIX daemons run with stderr connected to /dev/null.  Consider 
during development / debugging, or if you overlooked setting raiseExceptions 
altogether, or maybe you made the same mistake I did (because the 
reference-style documentation doesn't explain this behaviour).  My above 
proposal was to escape encoding errors only when raiseExceptions == False.  So 
when raiseExceptions == True, we might still lose log messages, and the 
UnicodeEncodeError might be printed to /dev/null i.e. also silently discarded.  
This suggests my proposal was trying to be too clever.

I guess the simplest approach is for emit() to handle encode errors by using 
'backslashreplace' to log the message AND calling handleError().  

Instead of the original code sketch, change StreamHandler.emit() to

        try:
            # issue 35046: merged two stream.writes into one.
            msg = self.format(record) + self.terminator
            stream = self.stream
            try:
                stream.write(msg)
                self.flush()
            except UnicodeEncodeError:
                # Try to log something, even pure mojibake might provide a clue
                encoding = getattr(stream, 'encoding', None)
                if encoding:
                    bytes = msg.encode(encoding, errors='backslashreplace')
                    msg = bytes.decode(encoding)
                    stream.write(msg)
                    self.flush()
                # Call handleError() as normal
                raise
        except RecursionError:  # See issue 36272
            raise
        except Exception:
            self.handleError(record)

(And I'd like a similar change for SyslogHandler at least).

----------

_______________________________________
Python tracker <rep...@bugs.python.org>
<https://bugs.python.org/issue36704>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

Reply via email to