Bug in base64, new module base32

[Gijs van Tulder]

Hi,

I'd like to send you two patches that are related to the base64 module. 
I hope you find them useful.

--

The first patch fixes a bug (or at least, I think it is a bug) in the 
base64_decode_alloc_ctx function. This method sometimes does not decode 
the final byte in the input. Strange enough, this bug was mentioned in 
the tests but not fixed.

Example (from the tests):

YWJjZGVmZ2hp   decodes to   abcdefghi

If you initialize a new decode-context buffer ctx and run:

base64_decode_alloc_ctx (&ctx, "YW", ...)
  ->  decodes nothing (that's OK)
      "YW" is kept in the decode-context buffer

base64_decode_alloc_ctx (&ctx, "JjZGVmZ2", ...)
  ->  decodes abcdef  (OK, that's "YWJjZGVm")
      now the decode-context buffer contains "Z2"

then, finally:

base64_decode_alloc_ctx (&ctx, "hp", ...)
  ->  decodes gh  (not OK, where's the final 'i'?)

I think this is because the base64_decode_alloc_ctx does not allocate 
enough space in the output buffer, because it doesn't look at what is 
already in the decode-context buffer. I found that if I add one extra 
byte to the buffer size it works. (See patch 1.)

--

The reason I found this bug was that I was working on a base32 module. I 
need a base32 encoding/decoding module, but Gnulib doesn't have one. In 
patch 2, I included a new module based on the base64 code that is 
already in Gnulib.

Kind regards,

Gijs van Tulder

0001-base64_decode_alloc_ctx-would-sometimes-miss-the-fin.patch
Description: Text Data

0002-New-module-base32.patch
Description: Text Data