Addresses some problems found while trying to use Pool::componentsByCategories

[aleixpol]

It was crashing because the list was not getting populated properly. This changes does so and addresses a couple of extra problems found while doing this as seen in their respective commits2.

[ximion]

The patches make sense to me except for the qt: Fix Pool::componentsByCategories one - which bug does that one fix?
Using g_new0 already ensures that the resulting memory is nulled, so no need to do that explicitly.
Did the strings from categories somehow have a shorter life than the surrounding container?

[aleixpol]

Did the strings from categories somehow have a shorter life than the surrounding container?

Note qPrintable is the result of .toUtf8().constData(). We could consider splitting it into two loops, one converting to QByteArray and the other creating the gchar **. We'd save a few memcpy() calls.

[ximion]

Sure, but does that mean the original bug was indeed the categories being destroyed before the function returned? That would be extremely surprising to me!
Using the deep-copy stringListToCharArray here is probably fine though, as the additional memory allocations won't cause a bug performance hit (the categories array is quite small, and the few nanoseconds saved for allocating & copying a few strings are likely unnoticeable. It's not like we are running this in a loop thousands of times)

[aleixpol]

If I don't have this commit I sometimes get this error message:
** (process:152147): WARNING **: 15:14:03.557: '' is not a valid XDG category name, search results might be invalid or empty.

I've changed the disputed commit to the slightly more verbose but correct and efficient version.

[aleixpol]

Ping?

[ximion]

If I don't have this commit I sometimes get this error message: ** (process:152147): WARNING **: 15:14:03.557: '' is not a valid XDG category name, search results might be invalid or empty.

Really strange! This does point at some kind of lifetime issue though.

I've changed the disputed commit to the slightly more verbose but correct and efficient version.

I would have even been fine with the previous version, but I do need to understand the rationale behind a change (merging blindly without understanding the cause will eventually haunt you).
But the new version is fine too :-)

Thank you for the patch!