Bug 105519 – Reducing print preview scale significantly causes a crash

• Component: Printing
• Status: RESOLVED FIXED

[Tim Horton]

The print preview image snapshot is being generated with a size that takes the print preview scale into account. It doesn't need to do this - the bitmap just needs to be the size that we're going to draw at!

<rdar://problem/12807090>

[Tim Horton]

Created attachment 180308 [details]
patch

[Alexey Proskuryakov]

Comment on attachment 180308 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=180308&action=review

> Source/WebKit2/ChangeLog:12
> +        The bitmap image used for print previews was being created respecting the print preview scale.
> +        This doesn't make any sense, as the scale does not affect the size of the image required to
> +        represent the previewed page. Instead, we should not scale the size, creating the buffer at a
> +        size that is constant regardless of scale, and do the scaling when drawing *into* the buffer instead.

With four lines of explanation, I'm still not quite sure if this was an out of memory crash :)

> Source/WebKit2/WebProcess/WebPage/WebPage.cpp:3283
> +        float printingScale = (float)desiredImageSize.width() / rect.width();

We generally prefer static_cast.

> Source/WebKit2/WebProcess/WebPage/WebPage.messages.in:220
> +    DrawRectToImage(uint64_t frameID, WebKit::PrintInfo printInfo, WebCore::IntRect rect, WebCore::IntSize desiredImageSize, uint64_t callbackID)

"Desired" makes it sound like the desire may not be granted. Is that accurate?

[Tim Horton]

(In reply to comment #2)
> (From update of attachment 180308 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=180308&action=review
> 
> > Source/WebKit2/ChangeLog:12
> > +        The bitmap image used for print previews was being created respecting the print preview scale.
> > +        This doesn't make any sense, as the scale does not affect the size of the image required to
> > +        represent the previewed page. Instead, we should not scale the size, creating the buffer at a
> > +        size that is constant regardless of scale, and do the scaling when drawing *into* the buffer instead.
> 
> With four lines of explanation, I'm still not quite sure if this was an out of memory crash :)

It was, but the crash was secondary to the fact that we were respecting the scale at all when we didn't need to :D

> > Source/WebKit2/WebProcess/WebPage/WebPage.cpp:3283
> > +        float printingScale = (float)desiredImageSize.width() / rect.width();
> 
> We generally prefer static_cast.

Of course!

> > Source/WebKit2/WebProcess/WebPage/WebPage.messages.in:220
> > +    DrawRectToImage(uint64_t frameID, WebKit::PrintInfo printInfo, WebCore::IntRect rect, WebCore::IntSize desiredImageSize, uint64_t callbackID)
> 
> "Desired" makes it sound like the desire may not be granted. Is that accurate?

No, it will always be the desired size. Perhaps requested? or just imageSize?

[Tim Horton]

http://trac.webkit.org/changeset/138356