Closed Bug 962546 Opened 12 years ago Closed 3 years ago

PDFs rendered slow with inappropriately high resolution [@ js::CompartmentChecker::fail(JS::Zone*, JS::Zone*)]

Categories

(Firefox :: PDF Viewer, defect, P3)

x86
Windows XP
defect

Tracking

()

RESOLVED WORKSFORME

People

(Reporter: rob1weld, Unassigned)

Details

(Whiteboard: [pdfjs-performance])

Attachments

(1 file)

185.86 KB, application/pdf
Details
STR: 1. View this PDF: http://www.ti.com/lit/ug/snau079/snau079.pdf 2. PgDN to Page 4 and notice how slowly the Page is being rendered; the Image's Background is being rendered with more detail than can be displayed on Screen. 3. Change the Zoom to "Page Width" and note that the details in the background do NOT become visible (all that rendering time bought you nothing). 4. While on Page 4 hit the 'Reload Circular-Arrow' to reload the Page and note the improved level of detail. You still do not get enough detail for how slow it renders but at least there IS some more details visible. It is at this point that the Browser crashed, I have this: bp- ad64fa02-0774-4f06-97ef-cbc922140122 . There is a Meta here: Bug 821733 .
I did not mark this a 'Blocker' for the 'Meta' because I can not reproduce the Crash; though the incorrect behaviour still counts as a Bug. During my testing I noticed that the 'Zoom' max. value was decreased from 400% to just 200%. I kept hitting the '+' to bring the Zoom up to %400 and re-rendered the Page. At 400% the Page has a significantly different aliasing or halftoning. When I decrease the Zoom back to "Page Width" the 'better?' (I guess it is correct) halftoning remains. When you reload the Page again (at "Page Width" Zoom) you lose the 'better' halftoning, and it is replaced by a somewhat similar rendering with those white blotches that appear regularly ('aliasing'). It looks and feels like we are going through two halftoning subroutines (accounts for slow speed and aliasing), and we are getting a 'different intent' displayed as a result (it looks wrong). --- I think this BR should go upstream. When I download the File and use Adobe Reader 11.0.06.70 to view it Page 4 takes nearly several seconds before the Background of that Image suddenly appears (presumably rendered correctly, it looks that way to me). Shortly after the Background is rendered the remainder of the Page is displayed quickly. Both our internal renderer and Adobe's Renderer have trouble with the 'PDF Code' used to create that texture.
Attached file snau079.4.pdf
Yes, this page #4 (joined to the bug) is very low to be fully rendered.
Confirmed in nightly 29.0a1 (2014-01-27), win 7 x64.
Status: UNCONFIRMED → NEW
Component: General → PDF Viewer
Ever confirmed: true
Please note this Bug is not about "slowness" of the PDF Viewer per se, but is about "suspected over-rendering" -- taking to much time on each 'portion' of the PDF and/or re-rendering and overlaying without what I would refer to as "Hidden Line Removal" (if this were 2D to 3D projection). We draw onscreen (instead of to a buffer, so it appears visually) and then draw over what we already drew (obscuring portions of it and taking a long time to calculate). This single Page also renders excessively slowly (and seems to not have enough detail for the time taken, thus I suspect oversampling here also) http://jmfriedt.free.fr/poster_12906.pdf . It takes so long it gives up and does not render the 'Data' within the smaller Table (next to the tiny Photo), that seems to be correctly rendered in Adobe Reader 11.
Priority: -- → P3
Whiteboard: [pdfjs-c-performance]
Here is another PDF that the Renderer gives up on: http://pro.jvc.com/pro/attributes/CAMERA/draw/gyhm850_890_dimensions.pdf . If you download that Link and use Acrobat to view it you can see a highly detailed drawing. If you view it in the Browser the drawing has NO detail at all, just an outline, and for what little is shown it clearly takes much too long to render.
Whiteboard: [pdfjs-c-performance] → [pdfjs-performance]
Severity: normal → S3

I just tested on Windows 11 and as far as I can tell, the performance issue has been fixed.
If someone can reproduce this issue, please reopen and share a profile.

Status: NEW → RESOLVED
Closed: 3 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: