KDE Gear projects with failing CI (master) (4 June 2024)

[Elvis Angelaccio]

On 09/06/24 17:16, Thiago Macieira wrote:
> On Sunday 9 June 2024 02:40:36 GMT-7 Elvis Angelaccio wrote:
>>> Yep. However, it's apparently considered an intentional behavior change by
>>> Qt. This also affected tests in other places (which meanwhile have been
>>> adjusted accordingly).
>>
>> That is surprising, isn't this kind of change forbidden in minor releases?
> 
> The path is correct at the point it is issued, referring to the same 
> directory. The fact that it is in a different form was not expected to be a 
> breaking change.
> 

I had a look at the 6.7 release notes and I've found the commit 935562a77b which seems to be
what changed this behavior.

Indeed, if I apply the same fix done in that commit to the test-case I attached before, 
i.e. if I replace

	auto tempDir = new QTemporaryDir(".test-qtempdir-");

with
	auto tempDir = new QTemporaryDir(QDir::currentPath() + "/.test-qtempdir-");

then the returned QTemporaryDir::path() goes back to being absolute.

I'd expect at least the QTemporaryDir documentation to get updated with a warning about this new behavior.
Currently it still claims that "Auto-remove mode will automatically delete the directory from disk upon destruction.", but this is no longer true
if you happened to create the QTemporaryDir with a relative template and then change the current directory in the process before deleting the QTemporaryDir instance.

Cheers,
Elvis