…rywhere rather than the XPI one

c_rehash is a compatibility perl script that isn't available on all
systems. In contrast, openssl rehash is available in more places, also
because c_rehash is sometimes implemented by using rehash.

Try to be more portable and use the right thing.

Now that the module may not be accessed directly, we may need to
redirect module stuff to the p11-kit module.

Since our add-on only cares about Belgian eID cards, ensure that if
another module is loaded through p11-kit, we filter those out by
skipping any slots that have a token but don't have "Belgium Government"
as the manufacturer.

We configure our PKCS#11 library into the p11-kit configuration, so that
applications (other than Firefox) can easily make use of our BeID
module.

In addition, we ship a Firefox add-on which uses the browser.pkcs11 API
to load our PKCS#11 module into Firefox automatically, and sideload that
extension.

Recent versions of Fedora have configured Firefox to autoload the
p11-kit-proxy.so module so that all registered PKCS#11 modules are
available on the system by default, which is great; but it does have the
unfortunate side effect that if our add-on tries to load the BeID
PKCS#11 module, that fails (because it has been loaded already).

Work around this issue by adding a configuration giving access to the
p11-kit-proxy.so module to our add-on; if (and only if) it fails to load
the BeID PKCS#11 module, we can then check if the p11-kit-proxy module
is loaded, and if so, assume that that was the reason and move on
without loading the module.

(cherry picked from commit db57066)

Recent mock packages don't install compilers anymore by default (?), so now we
need to depend on it...

Now that Firefox ESR 60 has been released, most distributions have
dropped ESR 52 for which the older XPC-based add-on was required.

The sole exception to this rule is RHEL6. One might wonder whether it's
worth keeping a package for a distribution that is that ancient; but
since it costs us next to nogthing, might as well keep the XPC extension
there.

Everywhere else, drop it and switch to the webextension.

…ed number

- Deduplicate the "log.h" filename, otherwise we can't correctly use it
  from some locations if we need to (this will break some builds...)
- Add a function to get a configuration value from the config file when
  using C code
- Add a function to figure out the current log level when using C code.

This will probably break the Windows and OSX builds; build system
updates will be necessary.

With that, PKCS#11 logging will work under Linux, even when we didn't
define _DEBUG or DEBUG.

And for keping debug and release behaviour similar, lets not use different init values for the debug level (in case the debug config is missing)

…nd set by VS by default for debug builds

We use DEBUG elsewhere, and _DEBUG here, for essentially the same thing.
That makes no sense. Fix.

The condition ((a != x) || (a != y)) is always true (it can't be equal
to two values at the same time). We actually meant to check if it is
equal to either of them, rather than to be not equal to both.

Coverity CID#191139

…nd them

…ller installer in order to have a signed wrapper around our uninstall script)