GitLab

committing at my discretion.

accidental remove like this.

[trac951] Revert "[trac951] we do have truncation support (at least in the api), so remove check and enable tests"

This reverts commit 76ae47d1.

We cannot yet remove the additional check and re-enable the corresponding
tests.  See my comment on ticket #951.

use "const" for second parameter (const size_t data_len).

This was noticed by Undefined symbol when linking when compiling
using Sunstudio.

Okayed on jabber.

the HMACImpl constructor.
okayed on jabber, directly pushing.

turns out botan had a check for exactly this, so we don't need to reinvent our own

Added separate checks for compatibility with 1.8 and 1.9, see configure.ac
cryptolink now has 3 special cases depending on version

with a longer length was wrong, even though it normally wouldn't cause visible
disruption and would return the expected result).

the case of output_len != sig_len.  Until we implement RFC4635 truncation
consideration won't work correctly (and could dangerously accept very
short sigs).  Since the truncation support is not so urgent, it would be
better to defer to it a seprate later task.

acked by shane on jabber

(with some non related cleanups)

also made the friend declaration more specific

and constify one thing

renamed the directory, libname, namespace and main class

- editorial/style fixes
- constify things in trivial cases
- catch exceptions by const reference