Fair point, one can set this in kdesrc-buildrc anyway.

This feels *very* special cased. Where does this automagic ends? Any why only in $HOME? We'll keep adding entries here forever... Why don't people use a .directory file for this?
If ~/Games is a de-facto standard, whoever creates it could add a .directory in there. If it's not, then that makes my point.

Very nice catch.

More casualties: the kpackage unittests...

It seems this commit is the reason for the unittest failure below?

Yep, tzset is really slow. I suppose there are other similar lines of code in KIO that would benefit from the same fix? KFileItem for example?

https://commits.kde.org/frameworkintegration/4cc12a5ad1fa92f8b934f02ff2e401539ec9b9f7

Ah, you need to specify that you don't have a developer account, otherwise the assumption is simply that you'll push the patch once approved.

(sure it's convenient for the unittest, but less meaningful for the users....)

Hmm hmm. The goal of this was to behave like "du" as much as possible. As you say, the real bug isn't that we count directories, but that we count files "as is" instead of blocks. But I'm not sure how easily we could get block information from here.

Cool, that debugging and profiling training I gave you last week is already paying off :-)

Nice catch.

Here's some debug output to understand what's going on:

I don't like magic based on process name. Just call a setter from the tests that sets a bool to enable the logging into m_testCookies? Inside the ifdef if you want, although the performance impact (of checking a bool) is negligible, so actually I'd remove the define/ifdef... (but I'm also OK with it if you want to keep it).

switch would require ugly fall-through, I prefer this code.

This map will keep growing and growing...

In D16509#352669, @dkurz wrote:

In D16509#352607, @dvratil wrote:

assuming DAV actually supports that.

It doesn't: https://stackoverflow.com/a/30871369.
I didn't find "color" nor "colour" in any CalDAV/WebDAV/iCal RFC either, except for one unrelated match in https://tools.ietf.org/html/rfc4791.

I'm having second thoughts about this one.

Laurent: can you do the merge to master? There's a whole lot of version numbers conflicts which I don't have time to sort out right now ;)

Don't show data source for child tasks, it's redundant.

Don't show project name for subtasks. Makes the view smaller indeed.

This seems copy/pasted from most of the attic cmake modules in ECM, which also say "the BSD license", with a reference to ECM's COPYING-CMAKE-SCRIPTS. Good enough?

Thanks!!

Use default-template-type trick, but it makes the additional "int" argument for the data() methods a bit weird (hidden magic)

Ah sorry I read too fast and I confused HKDFContext et BotanHKDFContext.

You're right. Fixed, thanks.