Bruce Momjian <bruce(at)momjian(dot)us> writes:
> I now understand the wisdom of your suggestion. Attached is a patch
> that removes hex_decode from ecpg properly, and returns -1 from the
> /common version.
I'm fairly unimpressed with this. I don't like having fundamentally
different (and 100% undocumented) behaviors between the frontend and
backend versions of the "same" function.
I think you should leave ecpglib alone. Unifying that little bit
of code isn't worth having to contort the API of the common version.
regards, tom lane