• Start dirlister on document signals

I'll certainly consider it.
Let's see if I can find the time. :-)

In D24881#553151, @bdevries wrote:

PS: Taking a step back, there are other ways of handling this. Other existing photo importers (across OSes) use different UI approaches. I've seen a few that handle this as part of the settings, offering a choice: "always keep", "always delete", "always ask". Others offer a checkbox "delete after import" on the photo import selection dialog.

That makes sense to me!

Swap keep and delete buttons. This makes keep the default action.

make sure that an url passed on the command line is loaded and shown before the possibly long running dirlister

Thanks!

In D24881#553151, @bdevries wrote:

Having said that, it could make sense to just swap the button around, but then I would say the two options clash with the original intent of the dialog box: "keep" would map to "yes" and "delete" would map to "no".

Add unit tests and solve bug when mulitiple files of same name exist.

Use QScopedPointer for QIODevice pointers

The purpose of this dialog box is to ask the user what should be done with the original files on the source device that the photos were imported from. I.e. deleting them on the source device, e.g. the camera or smartphone. Only the photos that were imported successfully are then deleted from the source device (so not the ones that failed or were not selected to be imported).
BTW, the dialog box pops up after the import itself has been done.

While this fixes the bug, I'm not quite sure it's the best way. Switching to KMessageBox::Dangerous seems inaccurate because even though there is indeed a dangerous action here, the dialog itself isn't about a dangerous action.

Don't worry, Nate will apply it.

Since this is my first ever contribution to KDE: is there anything that I still need to do on my side? Or will this commit be applied without further intervention from me?

Sounds lovely, please go ahead!

Sure, my "git" email address is devriesb@gmail.com .

Thanks! Can you provide your email address so I can commit this patch with correct attribution?

Searching and debugging and I just can't find the reason why.
So currently the counter works fine when the first image of a folder is opened and THEN one goes into fullscreen. But if one goes directly into fullscreen with the first image loaded, than it sets the counter to 0.

Removed comment that was used for the timer

Go ahead, take your time. :)

In D24622#548349, @ngraham wrote:

Did you mean to remove the timer in this patch?

Remove timer. Use simple method call instead.

Did you mean to remove the timer in this patch?

Reverted the timer to 0.
Now it works when switching to fullscreen. Next step: make the counter work properly when -f for fullscreen is used.

Revert timer and add additional updateDocumentCountLabel call

In D24622#547028, @tommo wrote:

And if updating the current index takes longer than 250 msec it would fail again, wouldn't it?

Probably, I can't get to that case where it's so slow. But I am aware that it can happen.

And if updating the current index takes longer than 250 msec it would fail again, wouldn't it?

I don't really understand. What Purpose dependency actually bothers you?

Something like this I think should work:

This results in the purpose functionality disappearing even though I have Purpose installed.

In D17732#545727, @sandsmark wrote:

In D17732#545644, @ngraham wrote:

Please feel free to send a Plasma patch!

That was the plan, but I don't have a dev environment for plasma set up, and submitting patches that don't even compile is a bit embarrassing. :-)

But fixing that should fix animations in other Qt apps with custom widgets as well.

Oh, I didn't know about that, thx for the hint. Will see how it can be integrated.

In D17732#545644, @ngraham wrote:

Please feel free to send a Plasma patch!

Any reason not to use kexiv2?

Please feel free to send a Plasma patch!

Something like this I think should work:

In D17732#545621, @sandsmark wrote:

In D17732#541174, @ngraham wrote:

In fact we now have a shiny new global animation slider in the General Behavior KCM. If you change the patch to make Gwenview respect that setting, I'll be happy with it.

Do you know what that adjusts? If it adjusts the normal Qt animation hints/properties like QStyle::SH_Widget_Animation_Duration this patch Should Just Work™.

In D17732#541174, @ngraham wrote:

In fact we now have a shiny new global animation slider in the General Behavior KCM. If you change the patch to make Gwenview respect that setting, I'll be happy with it.

Already fixed with another approach in master.

Wow, a very ambitious change.

In D24364#542199, @yurchor wrote:

Gwenview does not crash after the patch, but still animates the GIF's when "None" for animation is checked in the settings. Is this expected?

Gwenview does not crash after the patch, but still animates the GIF's when "None" for animation is checked in the settings. Is this expected?

In fact we now have a shiny new global animation slider in the General Behavior KCM. If you change the patch to make Gwenview respect that setting, I'll be happy with it.

In D17732#381774, @ngraham wrote:

However, I'm seeing that this appears to always disable all animations for me, no matter what value the animation speed is set up in the Breeze style settings:

In D17732#384423, @davidedmundson wrote:

As for the discussion above, I don't see how plasma-integration is relevant. It does control QStyleHints, but you're not using any of that.
Everything you use here comes from the QStyle which is all from Breeze.

I'm in China and the arcanist download takes forever and then fails.

Thanks, this is lovely. Much better. Can you provide an email address so I can land the patch with correct git authorship information?

Replaced the Label with a SpinBox

Removed the explicit call to GwenviewConfig::setJPEGQuality().
Still works.

On one hand, a slider looks nicer and it's easier to set gross values, but the lack of a text input field makes it harder to enter precise values. I wonder if we should replace the label that's next to the slider with an text field. That way we'd get the best of both worlds (and then we'd do this for Spectacle too).

Use y() + height() instead of bottom() + 1

I somehow knew autotest would fail on FreeBSD ... I cannot investigate why it fails, maybe too old MIME database?

The test fails here because the autotests/read/hdr/rgb.png file is 0 bytes, but if that is a phabricator artifact and it's really not a 0 bytes size, go ahead and commit :)

This task was briefly discussed at today's Dolphin BoF at Akademy. It was pointed out that the QGraphicsView dependency of the custom view-engine cannot go away, as that would not be worth the effort.

Add hdr to autotests

I'm going to land this now because I think the proposed new icon is sufficient. If we get a larger colorful version of view-preview we can always switch back to that.

Could you try to figure out if adding some autotest file would be hard?