test_cjose_concatkdf_derive_* fail on big endian architectures (s390x, powerpc, ppc64, sparc64, ...) about cjose HOT 4 CLOSED

@moschlar Re-reviewing the code with #38 in mind, the possible problem points I see are:

• Applying a 32-bit value to a byte buffer (including some pointer arithmetic)

  cjose/src/concatkdf.c

  Line 17 in b122181

  static uint8_t *_apply_uint32(const uint32_t value, uint8_t *buffer)

• More pointer arithmetic (via function calls)

  cjose/src/concatkdf.c

  Line 91 in b122181

  uint8_t *ptr = buffer;

• More direct pointer arithmetic while appending successive digests

  cjose/src/concatkdf.c

  Line 153 in b122181

  uint8_t *ptr = buffer + offset;

As with last time, we don't have a big-endian environment to troubleshoot and test on, so any assistance with or access to such an environment would be valuable.

from cjose.

• Applying a 32-bit value to a byte buffer (including some pointer arithmetic)

Ah yes, it seems to be called with value of anything but uint32_t:

• size_t:

  • cjose/src/concatkdf.c

    Line 35 in b122181

    ptr =_apply_uint32(len, ptr);

  • cjose/src/concatkdf.c

    Line 95 in b122181

    ptr = _apply_uint32(keylen, ptr);

• int:

  • cjose/src/concatkdf.c

    Line 140 in b122181

    _apply_uint32(idx, counter);

I'll see if I can somehow access such an environment to test it (cc @christmart).

Could you maybe think of a simple way to fix these three occurences?

from cjose.

(finally returning from holiday ... )

Thanks @moschlar.

The protocol expects a uint32_t, but just about all the calls around it expect a size_t. I'll change the int to uint32_t (it is just a counter that can't be negative). For the others I'll make an explicit cast.

from cjose.

The Debian architecture builds are now working, too! :-)

from cjose.