387018 - improve linux autoscroll icon

Status: RESOLVED FIXED

(Firefox :: Theme, defect)

Description

[dolphinling]

Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a7pre) Gecko/200707040404 Minefield/3.0a7pre

The autoscroll icon used to be transparent. When you had it over a bit of text, you could still see the text behind it. Sometime recently, this broke. 

Looking for a regression range now.

I find it hard to believe that I'm the only one that sees this, but I've looked and looked and no one's filed it yet.

Comment 1

[Adam Guthrie]

Probably due to bug 242621, but I've heard this is intentional.

Comment 2

[:Gavin Sharp [email: [e-mail]]]

Not so much "intentional" as a "known problem", I think. See bug 242621 comment 26.

Comment 3

[dolphinling]

Yeah, it's from 242621. I guessed that and verified it before getting distracted with important stuff for an hour.

I hope it's not intentional. I always considered it a very nice piece of polish, and the difference *is* noticeable (to me, at least).

Comment 4

[Chris Thomas (CTho) [formerly [e-mail] [e-mail]]]

This is intentional.  You can file a bug on making GTK not suck, if one is not already filed.  I know one is filed for Mac suckage.

Comment 5

[Daniel C]

I think a lot of people will notice this regression. Bug 242621 comment 30 mentions "trying to clean up any icon messiness that results". What exactly is the problem on Linux? Is there a GTK bug to point to?

Comment 6

[Michael Monreal [:micmon]]

Yes... this is very bad. If it is not possible to restore the FF2 look, then think about removing the graphic entirely. The non-AA/non-transparent icon currently used is terrible.

Comment 7

[[:Aleksej]]

To me, a bigger problem seems to be that it is not round. It is terrible that way together with opacity.

Comment 8

[Jesse Ruderman]

This was fixed weeks ago in bug 409478.  You guys can stop complaining now :)

Comment 9

[Jesse Ruderman]

Oops, this bug is for GTK, while that bug was for Mac.  I didn't realize GTK Firefox also had this problem.

Comment 10

[Ian Spence]

Attachment: My version of the autoscroll image

I gave up on this bug getting fixed for Firefox 3, so I just threw something together myself. It might not be polished enough for release, but if anyone thinks it is better than the current version, you can add it to your userchrome.css with

#autoscroller { 
  background-image: url("file://[PATH]/autoscroll2.png") !important;
}

Comment 11

[:Gavin Sharp [email: [e-mail]]]

If that image is better than what we have, why shouldn't we take it? What's not "polished enough for release" about it?

Comment 12

[Ian Spence]

I don't know, I only spent a couple minutes on it. Obviously it is free to use if others feel it is better

Comment 13

[dolphinling]

That image has semi-transparent antialiasing on the middle row of icons. AIUI, the middle row needs to have no semi-transparency (the right row can, the left row needs no transparency at all).

Still, if that works, then that proves that there aren't any problems with GTK, and the fix for linux should be equivalent to the one for OSX in bug 409478, right?

This could be a 6-line patch that there's already a reference for.

Or, I could have no idea what I'm talking about.

Comment 14

[Michael Monreal [:micmon]]

(In reply to comment #13)
> Still, if that works, then that proves that there aren't any problems with GTK,
> and the fix for linux should be equivalent to the one for OSX in bug 409478,
> right?
> 
> This could be a 6-line patch that there's already a reference for.

I had a small talk with some GTK guys because of this issue some time ago and the general answer was like "wtf"... They all said GTK was perfectly able to do this correctly, the only problem could be HOW mozilla tries to use GTK. Sadly, noone has stepped up to fix this yet, but since it is a quite visible regression, it would be nice to have this fixed for FF3 final.

Comment 15

[dolphinling]

Attachment: Changes to global.css and browser.xml - does not fix issue yet (but probably needed)

So I tried just copying the changes from the Mac patch; it didn't work. Apparently there's still more underlying work needed, like how bug 307204 was needed for mac before bug 409478 (the equivalent of this) was fixed. 

Still, I've attached my patch, since it'll probably be needed as the fix once GTK or our use of it or whatever is fixed.

Also, I was wrong about Ian's image above: the middle row doesn't use any semitransparency at all (as it shouldn't), so it is a definite improvement over what we have now. We should probably take it, but that would be a different bug.

Comment 16

[Reed Loden [:reed]]

Comment on attachment 301147 [details] [diff] [review]
Changes to global.css and browser.xml - does not fix issue yet (but probably needed)

>-              // Enable translucency on Windows and Mac
>-              this._autoScrollPopup.setAttribute("translucent", /Win|Mac/.test(navigator.platform));
>+              // Enable translucency on Windows, Mac, and Linux
>+              this._autoScrollPopup.setAttribute("translucent", /Win|Mac|X11/.test(navigator.platform));

Where are you getting X11 from?
javascript:navigator.platform returns "Linux i686" for me.

Comment 17

[dolphinling]

Oh, whoops! X11 is from navigator.appVersion, which is what the previous check is! navigator.platform gives "Linux i686" for me too, I'll rebuild and try with s/X11/Linux .

Comment 18

[dolphinling]

Attachment: Changes to global.css and browser.xml - does not fix issue yet (but probably needed)

Still doesn't fix the icon, but now this patch will work once whatever else needs to be done is done.

Comment 19

[:Gavin Sharp [email: [e-mail]]]

(In reply to comment #14)
> I had a small talk with some GTK guys because of this issue some time ago and
> the general answer was like "wtf"... They all said GTK was perfectly able to do
> this correctly, the only problem could be HOW mozilla tries to use GTK.

Yes, that's right. Who was blaming GTK? Most references to "GTK" in this bug are references to our GTK widget implementation. Someone needs to file and fix the Linux equivalent of bug bug 307204, as dolphinling points out, for this to be fixed as it was in 409478 for Mac.

(That being said, if we can get a better solution just by using a transparent PNG or something like that, we should do it. I don't know whether that will work.)

Comment 20

[Michael Monreal [:micmon]]

Did anyone already file this follow-up bug? If not, I can do it.

Comment 21

[dolphinling]

If anyone has, I haven't heard of it.

Comment 22

[Reed Loden [:reed]]

Hey guys... what's the latest on this?

monreal, if we need a new icon, can you get us one to use?

Comment 23

[Michael Monreal [:micmon]]

Reed, I'm unaware that anything changed and the code now supports a "nice" image for this. I fear this slipped through my hands, I thought I filed a followup for Linux but looks like I never did it :/ Is this still possible at this point?

Comment 24

[Ian Spence]

Attachment: Cleaner version of the autoscroll image

Nothing has changed to allow this to work with alpha-transparency, but I still think the current 1bit transparent image can be improved.

I believe it still might be beneficial to get a new image from an actual artist like jimmac.  The current file is a 3x3 grid that allows us to use the 1bit transparency now and alpha-transparency whenever it becomes available.

Also, marking bug 408284 as a dependency.

Comment 25

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Note that even when we implement alpha transparency for X/GTK, that will only work on systems using a compositing window manager (i.e., Compiz). The others will have to fall back to 1-bit transparency.

Comment 26

[u294409]

Robert, if everyone would think about older systems etc., then software development could stay still.

BTW, Metacity 2.22.0 has new compositor and can be used instead of Compiz. I don't see reason to prevent implementation of RGBA in Firefox.

Comment 27

[Ian Spence]

Notice that Roc said

>Note that even when we implement alpha transparency for X/GTK

He was not suggesting we not implement it, he was suggesting that we still think of the people (rough stat: pretty much every linux user) who doesn't use a compositor.

As Roc said in the other bug, the check for whether a compositor is running is extremely fast. Thus we should be able to use a system similar to how we check whether to show images in menu items to determine which image to show.

Comment 28

[Ian Spence]

Requested ui-review because while I know this isn't ideal, I believe it is far better than what we currently have

Comment 29

[Mike Beltzner [:beltzner, not reading bugmail]]

While I'm sure faaborg can do a good ui-r here, I think Michael Monreal or one of the guys who's been doing the icon work for gnomestripe shoudl weigh in.

To my eye, that's passable (and people say better than what you have now) so it's fine by me.

Comment 30

[Michael Monreal [:micmon]]

As bug 408284 won't be resolved for 3.0, I agree that we should use the new image Ian proposed. It's not the ideal solution but still better than the current image IMHO.

Comment 31

[Alex Faaborg [:faaborg] (Firefox UX)]

Comment on attachment 312582 [details]
Cleaner version of the autoscroll image

>It's not the ideal solution but still better than the
>current image IMHO.

I agree that it is better than the current images, but also agree that if one of the tango artists creates new images we should take those instead.

Comment 32

[Damon Sicore (:damons)]

Comment on attachment 312582 [details]
Cleaner version of the autoscroll image

a1.9+=damons

Comment 33

[:Gavin Sharp [email: [e-mail]]]

Which image is this supposed to replace? gnomestripe doesn't appear to have an existing autoscroll.png, so it looks to me like this requires packaging changes... but then if we're already using the Windows icon, why does it need to change?

Comment 34

[Ian Spence]

We've always used the same image as Windows. The difference is that linux users get to use the center column since we don't support alpha-transparent windows yet. I think it's best that we just add this icon to gnomestripe and edit jar.mn since we'll need to do it once this bug is truly fixed anyway, and I think it'd be too much of a hassle to deal with the Windows' daily redesign of icons anyway.

That being said, I'm winding down my senior year so if someone else could prepare the patch for the jar.mn, I'd be grateful

Comment 35

[:Gavin Sharp [email: [e-mail]]]

Attachment: jar.mn change

Comment 36

[:Gavin Sharp [email: [e-mail]]]

Checked in:
mozilla/toolkit/themes/gnomestripe/global/jar.mn 	1.40
mozilla/toolkit/themes/gnomestripe/global/icons/autoscroll.png 	1.1 

Can someone please verify this in the next nightly?

Comment 37

[Daniel C]

I see the updated icon, but this bug can't be "RESOLVED FIXED" under the current summary.

Comment 38

[:Gavin Sharp [email: [e-mail]]]

Despite the possible bug-hijacking that went on here, it would be best to file a new bug about any remaining needed work. Tracking multiple changes per bug (especially multiple changes made at different times and for different milestones) just leads to trouble.

I filed bug 429698.

Comment 39

[:Gavin Sharp [email: [e-mail]]]

(I would still appreciate confirmation that we did indeed improve the Linux autoscroll icon.)

Comment 40

[dolphinling]

I confirm the icon is improved. It's still imperfect (the edges look a bit flat) but it's much less jagged than before.

(And as the bug reporter, I approve of the hijacking/refiling) :)