With this new validation system each issue gets a long explanation as
well as a unique short tag which can be used for display it better in
UIs, ignore it or just quickly infer what the problem is.
This is a semi-breaking change as it will change the validator output,
so this will receive further updates in the next few days to make it
work better and display the output better.

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 79.2% (156 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

This is just to make naming consistent everywhere, in the code as well
as in printed messages.
For validation, severity == importance

This particular case of a component being an addon to itself is
nonsense, but people may make that mistake and we shouldn't crash in
that case.
With this patch the cache will be resilient against such cases and
simply ignore components depending on themselves.
We could still get nasty dependency loops though, with A depending on B
depending on A. This is a bit more complicated to resolve and will be
fixed in a future commit.
Resolves: #243

This should not fail, ever, unless we run out of memory. But since I was
looking at that code, having a sanity check here is better in case this
does become more relevant in future (and simply because not checking it
was not good prectice).

Currently translated at 12.7% (25 of 197 strings)

Ensure we never accidentally free unallocated memory.

This pretty much never happens on "normal" machines, but may be a thing
in empty chroots with special settings.

An error check was missing here. Thanks Coverity for pointing this out!

If the cache gets corrupted for whatever reason, AppStream shouldn't
hang in an infinite loop when trying to read it.

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

Resolves: #242

Currently translated at 100.0% (197 of 197 strings)

--install-docs=false can be used to disable installation of documentation,
which right now can't be disabled as it was possible with cmake.

Distributions like openSUSE don't ship an LMDB .pc file by default, and
since upstream doesn't offer the file there is a chance that other OSes
have similar issues.
So we look for LMDB without pkg-config as well now, to make building
AppStream easier on these platforms.

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

Since g_ptr_array_steal_index_fast() does not exist in older GLib
versions, we do something that is - sort of - equivalent when building
against older versions (but slightly slower and uglier).

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

Currently translated at 100.0% (197 of 197 strings)

This is the case in many automated test environments, e.g. when building
& testing Debian packages.
It also makes the cache more robust in general.