action #178855
openHandle broken unicode in files from wait_serial results
0%
Description
From #178597#21:
https://openqa.opensuse.org/tests/4912178#step/consoletest_finish/158 now shows: "Unable to decode consoletest_finish-160.txt."
The output in the file seems to be cut off, leading to broken unicode, so we cannot decode it.
The reason for that is the 4096 B ring buffer in serial_screen.pm's read_until
. By default, record_output => 0
, which means that on a match only 4096 B are returned, which might severe a UTF-8 sequence.
IMO either:
- Call
wait_serial
withrecord_output => 1
inread_until
- Try to keep UTF-8 intact (can happen on both ends of the buffer)
- Handle UTF-8 decode errors gracefully by using replacement characters instead of failing to decode everything
The current state means that the output is not visible without digging through the corresponding full serial log.
Files
Updated by tinita about 1 month ago
- Copied from action #178597: Various job pages do not load due JS error added
Updated by tinita about 1 month ago
- File consoletest_finish-160.txt consoletest_finish-160.txt added
- Subject changed from Display partially decoded content in wait_serial results to Handle broken unicode in files from wait_serial results
Attached file with broken unicode from the test
Updated by favogt 3 days ago
okurz wrote in #note-3:
@favogt @tinita can you please provide more context in the ticket,
Can't edit the description myself.
e.g. steps to reproduce
Something like script_run("systemctl status")
might already suffice there, with some luck there is a unicode sequence at end-4096B.
and impact on users
Under specific circumstances, the WebUI is unable to show the output of serial matches.