Unit test fixed. Unfortunately I can't avoid QFile::readAll() entirely. When the same file is saved as the one already open through mem-mapping, we are in trouble when writing to that file: "It is unspecified whether modifications made to the file made after the mapping is created will be visible through the mapped memory." Apparently this might be the case and causes strange behavior when working with the input data later on. And I can't simply unmap the input data, as it may be needed for transformation. Thus postpone QFile::readAll() to JpegContent::save(). At least it speeds up the thumbnail generator.

@matthieugras sorry this got lost. Once it's ready to land I can land it for you. While we're at it, would you like to try fixing blurry hidpi buttons too? See the suggestion at https://bugs.kde.org/show_bug.cgi?id=407726#c2

In D13095#467779, @romangg wrote:

In D13095#279659, @thsurrel wrote:

• accept to break some systems (but who wants that ?)

If these systems' breakage is because of an erroneous kernel driver, we can accept it. My system goes with brightness from 0 to 7500, and screen is activated at 1. It going dark at some higher value than 0 seems buggy to me. Is there some kernel directive on how it is supposed to work, i.e. if a display should be on at 1? @ngraham: what laptop model do you have?

Moving the brightness to zero is supposed to turn the backlight off, not the screen right?

• Add license headers

• use platformheaders

Fantastic. The dark inactive tab is maybe a bit too dark, but that's a fairly minor thing.

Thanks!

• Palette highligth fixes

This is fantastic. You fixed all the issues I was having, and the UX is practically perfect. The code looks totally sane too. In my testing I found only one small issue: when changing the icon size in one view and then switching to another view and then back, the first view did not save the new size; saving icon sizes only happens when closing the window. With more view modes prominently exposed now, it might be nice to save newicon sizes instantly, or when switching to a different view, rather than only saving when closing the dialog.

@leinir I have changed the output for the plasma tags. Please try, if it works for you now.

• Icon / Tooltip

If the maximize button is a toggle, it needs to look visually toggled when active. Also the icon should be something more like view-fullscreen.

• Fix build

• Add toggle of the Maximize / Minimize split

In D21312#467791, @lavender wrote:

In D21312#467698, @ngraham wrote:

+1 for restoring the feature, which is very useful! I've wanted this myself many times, so thanks for submitting this! However I'll admit I'm not super thrilled about having the capacity indicator bar always visible. I can't quite put my finger on why, but something about it feels a bit off. Maybe that's why the original only appeared on hover? But that's problematic in an era where we need to be designing for touch-friendliness. Maybe it could be a pie chart to the right of the icon but left of the text? Or maybe it's fine as-is and I'm just being a stick-in-the-mud.

Either way, adding VDG for comment.

I agree that there's something off about it - to me it is the fact that it's a lot of visual clutter that does not give actionable information. It relies on the user knowing how big the volumes are because for example a 90% filled bar is very different to someone with a 100 GiB volume than someone with a 1TiB volume. I don't have any data on this but I have a feeling most people don't care about how full all their volumes are until it hinders them so how I suggest we do it is for it to stay out of the user's way until the volumes are almost full, when there is a reason to present the information and possible actions the user can take.

Or at least vastly reduce how much attention it grabs because it could just be me but it distracts from reading the volume labels, I imagine it can get draining if you have to scan through a lot to find the one you want.

+1

The previous made changes are reverted and the faulty method declaration is removed.

I agree with you, that this method can be removed.
So I will modify the review accordingly.

In D21301#467903, @tcanabrava wrote:

@hindenburg I have a long list of branches that are build on top of other branches, and phabricator is breaking my reviews (four to date, just this week).
would you accept konsole to move to gitlab?

It's on my to-do list today. I took a look and here are some initial comments:

• The button next to the close button needs a tooltip because its icon doesn't make it obvious what it does. Maybe a better icon too, like zoom-in or view-fullscreen
• Once in zoomed/maximized mode, it's not obvious how to get out of it
• Might wanna add a horizontal line under the menubar to separate it from the split views' headerbars

@hindenburg I have a long list of branches that are build on top of other branches, and phabricator is breaking my reviews (four to date, just this week).
would you accept konsole to move to gitlab?

phabricator is tracking the wrong branch , it's *not* merged. (can I have a review? :)

Hmm, I wonder how this fixes the crash, but WebEngine is magic (plus this looks "harmless")

In D16648#467854, @fvogt wrote:

It looks like several QUrls were changed to QString here, why? Most of those QStrings, if not all, actually contain URLs.

• kpackage as a qt resource (that Marco started):

That was blocked after fvogt complained about it saying he's sure it would have the exact opposite effect (slowing startup time and increasing memory usage).
Now I don't know what would really be the difference, on the tests i did i failed to find any appreciable difference in any direction neither in startup time (with aleix startup tracking class) nor in memory usage, I don't know how to do really precise measurements, but seems overall pretty non influent

I think this is fine, and an improvement.