Bug #21718
Updated by Peter Amstutz 4 months ago
Unfortunately there seems to be a regression in the 2.7.2 release that wasn't caught by our tests.
What seems to have happened is that the data type returned by the Python SDK when reading from Keep changed to avoid copying a buffer (as an optimization), but the methods available on the new type are slightly different. Specifically throwing an exception on the call to "decode()":
<pre>
if "cwl.output.json" in outc:
with outc.open("cwl.output.json", "rb") as f:
if f.size() > 0:
outputs = json.loads(f.read().decode())
</pre>