What happened?
_read_binary_value's fallback calls base64.b64decode(val) without validate=True. Since b64decode ignores non-alphabet characters by default, many ordinary strings "successfully" decode into garbage bytes, so the intended UTF-8 fallback rarely fires, a plain-text column value can be turned into garbage and sent as image/audio bytes.
File: .../huggingFace/codegen/HuggingFaceCodegenBase.scala (_read_binary_value). Fix: pass validate=True and catch binascii.Error so only genuine base64 decodes.
How to reproduce?
Put a non-data-URL, non-base64 string in an image/audio column -> it's decoded to garbage bytes instead of being treated as text.
Version/Branch
1.3.0-incubating-SNAPSHOT (main)
Commit Hash (Optional)
No response
What browsers are you seeing the problem on?
No response
Relevant log output
What happened?
_read_binary_value's fallback callsbase64.b64decode(val)withoutvalidate=True. Sinceb64decodeignores non-alphabet characters by default, many ordinary strings "successfully" decode into garbage bytes, so the intended UTF-8 fallback rarely fires, a plain-text column value can be turned into garbage and sent as image/audio bytes.File:
.../huggingFace/codegen/HuggingFaceCodegenBase.scala(_read_binary_value). Fix: passvalidate=Trueand catchbinascii.Errorso only genuine base64 decodes.How to reproduce?
Put a non-data-URL, non-base64 string in an image/audio column -> it's decoded to garbage bytes instead of being treated as text.
Version/Branch
1.3.0-incubating-SNAPSHOT (main)
Commit Hash (Optional)
No response
What browsers are you seeing the problem on?
No response
Relevant log output