656284 - XUL progressmeter is clipped on Windows

Status: VERIFIED FIXED

(Core :: Widget: Win32, defect, P1)

Description

[Jim Mathies [:jimm]]

Attachment: download progress

STR:

1) open download window
2) download something

see screenshot.

Looks like we are going to have to track down that weird lower bounds. (kProgressDeterminedXPOverflow is the patch in bug 568825)

Comment 1

[Mounir Lamouri (:mounir)]

This is really weird because, indeed, there is an issue with the download window but also, on my Windows VM, I don't have any overflow now (that means all progress bars look like the download window progress bars) but I do on Nightly.

Comment 2

[Chris B.]

This also occurs in the very quick pop-up for the "Processing Print" submission dialog.

Comment 3

[Mounir Lamouri (:mounir)]

Actually, it seems to happen with all XUL progressmeter element.

Comment 4

[Raul Nicolae Malea]

Also in progress add-ons verify after change Firefox 4 to Nightly.

Comment 7

[Mounir Lamouri (:mounir)]

Attachment: Patch v1

Roc, this patch seems to fix the issue though I still don't understand why the issue is only showing on Windows.

Comment 8

[Mounir Lamouri (:mounir)]

We really need this to be fixed for Firefox 6 branching. Otherwise, <progress> will probably miss the Firefox 6 train :(

Comment 9

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

Can you explain why you're making these changes? I don't understand what the bug is or how these changes fix it.

Comment 10

[Mounir Lamouri (:mounir)]

(In reply to comment #9)
> Can you explain why you're making these changes? I don't understand what the
> bug is or how these changes fix it.

It's a regression from bug 568825. I have to admit I don't really understand what is happening but basically, there was this overflow that I thought was a bug somewhere between our Windows widget and Windows. So I wrote a dirty workaround that happens to broke XUL progressmeter.
It seems patch fixes the issue for HTML progress elements and doesn't regress XUL progressmeter. I'm not sure if that's a real fix but it seems to work and is obviously better than the previous fix.
I believe you have better knowledge of layout/ and widget/ code that me so you might understand better than me what's happening...

Comment 11

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

Comment on attachment 532156 [details] [diff] [review]
Patch v1

Review of attachment 532156 [details] [diff] [review]:
-----------------------------------------------------------------

::: layout/forms/nsProgressFrame.cpp
@@ +173,5 @@
>  {
>    bool vertical = GetStyleDisplay()->mOrient == NS_STYLE_ORIENT_VERTICAL;
>    nsHTMLReflowState reflowState(aPresContext, aReflowState, aBarFrame,
>                                  nsSize(aReflowState.ComputedWidth(),
> +                                       aReflowState.ComputedHeight()));

Setting an available height seems wrong for the reasons we discussed in the other bug.

Comment 12

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

I don't understand what these overflow values are supposed to mean.

Comment 13

[Mounir Lamouri (:mounir)]

Attachment: With -moz-box-sizing

While looking around the code I realized that "-moz-box-sizing: border-box;" was used to prevent some overflows. I don't know if not using that was a real mistake but it seems to fix the issue for vertical and horizontal progress bars on Windows and do not regress anything on GTK.

Roc, does that sound like a real fix or still a workaround? If the later, is it safe?

Comment 14

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

Comment on attachment 532450 [details] [diff] [review]
With -moz-box-sizing

If this works, great.

Can you make a reftest?

Comment 15

[pal-moz]

Attachment: screenshot

on Windows XP, meter is blank/empty.


GPU Accelerated Windows : 0/1

Comment 16

[Mounir Lamouri (:mounir)]

Attachment: With tests update

I had a reftest checking that there were no overflow. I guess I can keep it. To test the XUL regression, the only idea I have is to compare a very small <progresmeter> with a <progressmeter> styled so there is no bar (like done for <html:progres> in layout/reftests/native-theme/). Though, I don't know how to do that in XUL (can I access an anonymous element from a reftest?).

Please, let me know if you see a way to test that.

Comment 17

[Mounir Lamouri (:mounir)]

Pushed:
http://hg.mozilla.org/mozilla-central/rev/93656440018a

Comment 18

[Mounir Lamouri (:mounir)]

Can anyone confirm that the bug is fixed in the current Nightly and mark the bug as VERIFIED?

Comment 19

[Chris B.]

(In reply to comment #18)
> Can anyone confirm that the bug is fixed in the current Nightly and mark the
> bug as VERIFIED?

Nightly isn't out yet - but I will confirm when the build is available to download.

Comment 20

[Chris B.]

PARTIALLY confirmed.  There is still some extra white space on the bottom of the progress bar about a couple pixels wide, but it still looks better than the original screenshot above "download progress".  Still needs a slight modification, though.....

Comment 21

[Chris B.]

Attachment: Download Window (5-17-11)

Comment 22

[Chris B.]

(In reply to comment #21)
> Created attachment 532950 [details]
> Download Window (5-17-11)

Sorry....wrong upload.....

Comment 23

[Chris B.]

Attachment: Download Window (5-17-11)

Comment 24

[Josh Tumath]

The gap appears all the way around the bar, as though it has 1px padding. It's just more noticeable at the bottom.

Comment 25

[Chris B.]

(In reply to comment #24)
> The gap appears all the way around the bar, as though it has 1px padding.
> It's just more noticeable at the bottom.

Right - but I believe if the progress bar is say 20 pixels high, the entire 20 pixels should be filled in, like with the 5/10/11 build (last working one) and not only 14 or 16 pixels centered throughout.

Comment 26

[George Carstoiu]

Attachment: Screenshot after 2nd patch

Mozilla/5.0 (Windows NT 6.1; rv:8.0a1) Gecko/20110719 Firefox/8.0a1

Verified issue on the latest trunk. Adding a screenshot of how the download progress bar looks in the latest Nightly.

Comment 27

[George Carstoiu]

Mozilla/5.0 (Windows NT 6.1; rv:6.0) Gecko/20100101 Firefox/6.0

Verified issue also on 6.0b2 - not reproducible.

Comment 28

[Wesley Johnston (:wesj)]

How was this verified fixed if there are still issues here?

Comment 29

[George Carstoiu]

Looking at the latest trunk I haven't noticed any issues. Can you please elaborate on what issues are you still observing?

Thanks!

Comment 30

[Chris B.]

(In reply to comment #29)
> Looking at the latest trunk I haven't noticed any issues. Can you please
> elaborate on what issues are you still observing?
> 
> Thanks!

As stated in comment 20 two months ago, the green fill in does NOT fill the entire rectangle when a file is downloading.  This also appears during the "Print Preparation" dialog.  I still see this issue in the latest Nightly.  Please refer to the attachments above:
  1) Download Window (5-17-11)
  2) Screenshot after 2nd patch

Comment 31

[Philip Chee]

Please provide a new screenshot using the latest nightly Firefox 8.0a1

Comment 32

[Peter Henkel [:Terepin]]

This bug is there in the same form from the day this bug landed until today. What's the point of posting another screenshot showing the same thing? You think we are making this up just to make you mad or something?

Comment 33

[Mounir Lamouri (:mounir)]

This bug is fixed, the remaining problem will be treated in bug 658885.