Matthias Klumpp

That's fine then :-) - It's a nice thing if upstream URLs are also HTTPS (especially if usually both variants are provided and sometimes a http link is an accident),...

> Using compiler directives to export symbols from shared builds via defining macros. Can this be done differently? I really don't like this, and it adds a ton of unnecessary...

To elaborate on the exported symbols: AppStream exports everything on Linux/FreeBSD/MacOS that is not wrapped by `AS_BEGIN_PRIVATE_DECLS`/`AS_END_PRIVATE_DECLS` which simply switches the default symbol visibility to "hidden". Stuff in `*-private.h` headers...

It's a bit ugly, but I guess we need to accept that :-) Why is the CI failing though?

Hmm, I'd need to learn more about this first to properly judge it, but first of all thank you for the patch! Using `QUtf8StringView` sounds attractive because it makes it...

Looks like we do allow digits just fine, but if the first character is a dash, dot or any other such thing this hint is generated. Starting the main description...

Using `QUtf8StringView` and copy in case there is no NULL-terminator indeed seems like the best idea, since QString is UTF-16, so when passing that we would need to copy anyway....

Do you think we can get this resolved for the AppStream 1.0 release? (Intended to be released in ~2 months)

Also, how would this be implemented if we had a Qt5 version compiled as well? (Most likely that will happen, to make the transition to AppStream 1.0 easier).

> But the issue there is `id` is more reliable and icon tags may not always exist. If there is an icon, and icon tag will also exist, so you...