Closed
Bug 1285263
Opened 8 years ago
Closed 8 years ago
Parts of text are not displayed
Categories
(Core :: Graphics, defect)
Tracking
()
VERIFIED
FIXED
mozilla50
People
(Reporter: roxana.leitan, Assigned: nical)
References
()
Details
(Keywords: regression)
Attachments
(2 files)
305.78 KB,
image/png
|
Details | |
1.07 KB,
patch
|
bas.schouten
:
review+
|
Details | Diff | Splinter Review |
20160706030233 Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:50.0) Gecko/20100101 Firefox/50.0 [Affected versions]: Nightly 50.0a1 [Steps to reproduce]: 1. Open FF with Clean Profile 2. Open page: https://imagej.nih.gov/ij/docs/pdfs/examples.pdf 3. Press F5 to reload the page [Expected result]: Page content is correctly displayed [Actual result]: Parts of text are not displayed [Regression range]: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=57cf7cae92f1b234b763240a33f030fd68fc2166&tochange=56715a33b016d05afc2541e19fd7e0c907258971 [Additional notes]: Reproducible also on Mac OS X 10.10
Comment 1•8 years ago
|
||
Removing the dependency on bug # 1282871 as this appears to be a rendering issue rather than an issue relating to the Family Safety root work under Win 8.1. The above regression range points to bugs that are related to rendering/canvas.
No longer depends on: 1282871
Comment 2•8 years ago
|
||
[Tracking Requested - why for this release]:
Blocks: 1167235
tracking-firefox50:
--- → ?
Component: PDF Viewer → Graphics
Flags: needinfo?(nical.bugzilla)
Flags: needinfo?(bas)
Keywords: regression
Product: Firefox → Core
Updated•8 years ago
|
OS: Windows → All
Assignee | ||
Updated•8 years ago
|
Assignee: nobody → nical.bugzilla
Flags: needinfo?(nical.bugzilla)
Updated•8 years ago
|
See Also: → https://github.com/mozilla/pdf.js/issues/7486
Assignee | ||
Comment 5•8 years ago
|
||
Fix by the latest patches in bug 1167235. Please reopen if it reproduces again.
Status: NEW → RESOLVED
Closed: 8 years ago
Flags: needinfo?(bas)
Resolution: --- → FIXED
Assignee | ||
Comment 6•8 years ago
|
||
Actually, I can't reproduce this on an optimized build but I it still reproduces with a debug build, so there seem to be a timing thing to figure out still.
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Comment 7•8 years ago
|
||
I can still reproduce the problem on Latest Nightly50.0a1[1]. [1] https://hg.mozilla.org/mozilla-central/rev/0fbdcd21fad76a00328e67875c6f40dc219235f4 Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:50.0) Gecko/20100101 Firefox/50.0 ID:20160718030454
Comment 8•8 years ago
|
||
(In reply to Nicolas Silva [:nical] from comment #6) > Actually, I can't reproduce this on an optimized build but I it still > reproduces with a debug build, so there seem to be a timing thing to figure > out still. Note also the intermittent failure filed as bug 1284523, which I'm guessing may be the same underlying problem -- and its intermittent nature suggests timing may be involved.
Assignee | ||
Comment 9•8 years ago
|
||
The issue is caused by clips being saved or restored improperly (if I remove all clips the issue disappears) on the DrawTarget. Still digging...
Assignee | ||
Comment 10•8 years ago
|
||
Ha! Clips are affected by the current transform which means we have to restore them using the corresponding transform in the style-stack, and not the final transform. If the succession of SetTransform calls happens to be expensive, we can think about maintaining a separate stack of clips manually transformed in the final transform when we return the DrawTarget, but I'd rather not add this kind of trickiness if we can avoid it.
Attachment #8772837 -
Flags: review?(bas)
Updated•8 years ago
|
Attachment #8772837 -
Flags: review?(bas) → review+
Comment 11•8 years ago
|
||
(In reply to Nicolas Silva [:nical] from comment #10) > Created attachment 8772837 [details] [diff] [review] > Use the proper transform when restoring clips > > Ha! Clips are affected by the current transform which means we have to > restore them using the corresponding transform in the style-stack, and not > the final transform. > > If the succession of SetTransform calls happens to be expensive, we can > think about maintaining a separate stack of clips manually transformed in > the final transform when we return the DrawTarget, but I'd rather not add > this kind of trickiness if we can avoid it. Good catch, quite a stupid mistake. Pretty sure I wrote that code. Can you add a test?
Comment 12•8 years ago
|
||
Pushed by nsilva@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/fff92b2a76ec Restore clips with the proper transform when in CanvasRenderingContext2D::EnsureTarget. r=bas
Assignee | ||
Comment 13•8 years ago
|
||
(In reply to Bas Schouten (:bas.schouten) from comment #11) > Good catch, quite a stupid mistake. Pretty sure I wrote that code. Can you > add a test? Will do.
Comment 15•8 years ago
|
||
bugherder |
https://hg.mozilla.org/mozilla-central/rev/fff92b2a76ec
Status: REOPENED → RESOLVED
Closed: 8 years ago → 8 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla50
Reporter | ||
Comment 17•8 years ago
|
||
Verified as fixed on Windows 10 and Mac OS X 10.10 on the latest Nightly 50.0a1(Build ID: 20160724030208) - Page content is correctly displayed
Status: RESOLVED → VERIFIED
You need to log in
before you can comment on or make changes to this bug.
Description
•