KDE Frameworks with failing CI (master) (5 January 2024)

[Sune Vuorela]

On 2025-01-07, Harald Sitter <sitter at kde.org> wrote:
> Something changed server side? Although our code is also meh.

I asked around and apparantly some things changed. From the dev on the
other end:

| Ok, maybe I understand the problem. Until now, users of our API could
| call up an infinite number of pages when they searched for something.
| (And they did under certain conditions) The API always responded with a
| blank page. 
|
| From my point of view, this made no sense at all, because the total
| number of hits is returned in the meta information and also the number
| of elements on a page, which makes it easy to calculate max. pages.
| The API now treats it as a faulty request if a page outside the result
| set is requested. Also, the absolute values of integer values are now
| always used - i.e. unsigned.
|
| I'm happy to discuss if a different handling is desired.

I guess, do any of our 'real code' actually end up sending -1 to the
server, or is it just the unit test?

If real code does it in practice, maybe we should convince the store
people from the other end to special case (unsigned int)(-1) and map it
to 0 for the page argument for a year or two while we do the
std::max(page,0) that Harald was suggesting and get it to distributions.

If it only is in the unit tests, I guess we should just change them to
the std::max(page,0) that Harald suggests.

/Sune