encoding: inconsistent panic behavior for {hex,base32,base64}.Decode #58391
Description
CL 461958 introduced inconsistencies across two dimensions:
Encodestill panics on output buffers too small, whileDecodenow treats such cases as an error. This inconsistency cannot be resolved sinceEncodedoes not report an error and could be an argument for why we should revert CL 461958.- The
base32andbase64packages share very similar API ashex, and their equivalentDecodefunctions still panic if the output buffer is too small. We should handle these cases in identical ways across packages.
Should we revert CL 461958 to maintain consistency between Encode and Decode?
Or should we roll forward with applying equivalent checks to base32 and base64?
I have a very minor preference for rolling back.
Related: the fact that Encode and Decode can panic is a sharp edge fixed by #53693.