190255 - Off-by-one pixel when scrolling vertically

Status: RESOLVED FIXED

(Core Graveyard :: GFX: OS/2, defect)

Description

[Jim Moe]

User-Agent:       Mozilla/5.0 (OS/2; U; Warp 4.5; en-US; rv:1.3a) Gecko/20021210
Build Identifier: Mozilla/5.0 (OS/2; U; Warp 4.5; en-US; rv:1.3a) Gecko/20021210

Text: When scrolling up or down, the renedered text can sometimes be missing a
line one pixel thick, or have an line one pixel duplicated.
Images: The appearance of one or more horizontal white lines one pixel thick.
Re-load: The page usually scrolls up (towards the top pf a page) one pixel when
a page is re-loaded. This does not always happen. (Obviously not when at the top
of a page.)

Reproducible: Always

Steps to Reproduce:
1. Pick a site
2. Scroll down a little bit.
3. Scroll up.

Actual Results:  
  Some lines of text have an extra line in them making the line look taller than
others. Other lines have a line missing.
  Some images have white horizontal lines. The placement and quantity of the
lines varies.

Expected Results:  
No extra or missing lines.
The rendered page should not shift vertically when it is re-loaded and the
content has not changed.


  I do not recall when I first noticed this; somewhere around v1.0, maybe
earlier. It has been common in v1.1 and v1.2, less so in v1.3a.
  Selecting the mis-rendered text forces a re-rendering which corrects the
blemish. Covering and uncovering an image gets rid of the white lines.
  The problem is more likely to occur toward the bottom of the page.

Comment 1

[Jim Moe]

Attachment: Example of text with a missing line.

Comment 2

[Jim Moe]

Attachment: Example of an image with white lines.

Comment 3

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

This could be an OS/2 only issue. I haven't seen this on Linux for a long time.

Comment 4

[Jim Moe]

Attachment: Second example of off-by-one

Another aspect of this defect is how tables are displayed. Every third or
fourth line has an extra space in it.
  The left half of the image shows the re-paint after that section had been
covered by another application. The right half is how it was originally
displayed (and desired). It is normal for the converse to occur.

Comment 5

[Tom Mraz]

*** This bug has been marked as a duplicate of 171282 ***

Comment 6

[Jim Moe]

  The text aspect of this problem has a solution: install the Freetype font
smoothing software (technology preview) offered by Innotek (www.innotek.de).
Besides doing a very nice job of artfully blurring text, the problems I had with
font rendering have completely disappeared.
  Another benefit relates to scrolling with a mouse. Scrolling down would
initially have a jump up before continuing down. The further down a page, the
larger the jump. This behavior has disappeared as well.

Comment 7

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Reopening.

Comment 8

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Moving to Gfx:OS/2.  I suspect the main problem is this commented-out code in
nsDeviceContextOS2::Init, which should replace the code that's uncommented:

  // make p2t a nice round number - this prevents rounding problems
//  mPixelsToTwips = float(NSToIntRound(float(NSIntPointsToTwips(pt2t)) /
float(dpi)));
//  mTwipsToPixels = 1.0f / mPixelsToTwips;
  mTwipsToPixels = ((float)dpi) / (float)NSIntPointsToTwips(72);
  mPixelsToTwips = 1.0f / mTwipsToPixels;

This would certainly cause problems of this type.

Comment 9

[Mike Kaply [:mkaply]]

Attachment: Fix

I agree with dbaron. We should be rounding.

I also took this opportunity to cleanup the corresponding code in printing as
well.

Comment 10

[Mike Kaply [:mkaply]]

Comment on attachment 130782 [details] [diff] [review]
Fix

sr=blizzard (platform specific)
a=mkaply

Comment 11

[Mike Kaply [:mkaply]]

Fixed on trunk and branch

Comment 12

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

ahm, this could break printing on OS/2 a la bug 211092

Comment 13

[Mike Kaply [:mkaply]]

I'll do some more testing tomorrow.

It's easy not to do the rounding for printing on OS/2 since we follow different
codepaths.

Unfortunately I compared my reslts to Windows to see if I was printing correct.
Oops :)

Comment 14

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

Just test the testcases in bug 211092 to make sure that 100mm boxes print as
100mm, etc.

Comment 15

[Mike Kaply [:mkaply]]

You are correct sir.

I took out the rounding in the printing section.

Comment 16

[Joosep-Georg Järvemaa]

This is dupe of bug #94739 and is NOT FIXED, I think...