utils: run() handle unicode decode failure #132
Conversation
Expand the run() wrapper to handle logging and unicode decode failures in general. LP: #2011344
|
I'm a bit confused, does this change solve a problem or does it give more information to investigate? I wonder if we should not define a error handler to log the error and then call excerpt import codecs
def f(exc):
print("oops", exc)
return codecs.backslashreplace_errors(exc)
codecs.register_error('probert-run-unicode', f)
data = b'r\xe9serv\xe9e'
print(data.decode(errors='probert-run-unicode'))(we can pass the handler to subprocess directly) |
dumpe2fs is passing the latin1 string, which causes problems when we try to act on the non-label output. I'll incorporate your suggestion, which I think salvages the sizing information we are intrested in. |
The goal is a bit of both. Also, the goal is to log the reasons why the resize tool didn't succeed, so it shows that as well. |
| for line in stream.splitlines(): | ||
| for line in log_input.splitlines(): | ||
| log.debug(line) | ||
| return text |
There was a problem hiding this comment.
if we still return None here in case of UnicodeDecodeError, we don't solve the issue with the sizing information, do we?
2 participants
Expand the run() wrapper to handle logging and unicode decode failures in general.
The original problem reports udev decode failures involving a NTFS partition with a label similar to 'NTFS réservée'. I am unable to construct that with NTFS, but a ext4 version was instructional. The subsequent dumpe2fs has invalid output.
Pyudev seems to handle this all well, I was unable to coax decode failures from it.
I think the robustness check is interesting for now, and if we get more info on reproduction steps, further work can be considered.
LP: #2011344