Bug 54420

Summary: SharedBuffer::buffer() misses purgeable data, sometimes breaking re-decodes on macs.
Product: WebKit Reporter: Gavin Peters <gavinp>
Component: Page LoadingAssignee: Nobody <webkit-unassigned>
Status: RESOLVED FIXED    
Severity: Normal CC: commit-queue, gavinp, ggaren, koivisto, pkasting, sam, simon.fraser
Priority: P2    
Version: 528+ (Nightly build)   
Hardware: PC   
OS: OS X 10.5   
Attachments:
Description Flags
Patch none

Gavin Peters
Reported 2011-02-14 16:02:26 PST
On the mac platform, when we build with ENABLE(PURGEABLE_MEMORY), the call SharedBuffer::buffer() can return an empty vector when the SharedBuffer contains no data, but has purgeable memory associated with it. Some image decoders (JPEGImageDecoder & WEBPImageDecoder) rely on buffer() to get their data, and as a result are actually broken after the original image data has migrated to purgeable memory. This migration occurs on Images in CachedImage::destroyDecodedData(), which makes JPEG and WEBP images unreconstructable by the JPEGImageDecoder after we've destroyed the decoded data, a very undesirable result, as subsequent decode attempts get a 0,0 image.
Attachments
Patch (4.67 KB, patch)
2011-02-14 16:35 PST, Gavin Peters 		no flags
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
Gavin Peters
Comment 1 2011-02-14 16:35:40 PST
Created attachment 82382 [details] Patch
Peter Kasting
Comment 2 2011-02-14 16:38:27 PST
Comment on attachment 82382 [details] Patch Image parts LGTM
Gavin Peters
Comment 3 2011-02-14 16:40:38 PST
Comment on attachment 82382 [details] Patch This patch fixes the issue by removing consumers of SharedBuffer::buffer() and using the safer SharedBuffer::data(). pkasting pointed out (in irc) that there may be further simplification to be had by making all decoders use the same interface. I believe that! However this bug was causing chrome to stop displaying images on many sites, and so I made this minimal fix right away. I'm open to ideas on testing, and also to follow up with more rationalization of how we handle SharedBuffer. See: http://code.google.com/p/chromium/issues/detail?id=68622 for the chrome bug that brought me here.
Antti Koivisto
Comment 4 2011-02-15 01:26:27 PST
Comment on attachment 82382 [details] Patch r=me
WebKit Commit Bot
Comment 5 2011-02-15 05:07:33 PST
The commit-queue encountered the following flaky tests while processing attachment 82382 [details]: http/tests/websocket/tests/multiple-connections.html bug 53825 (author: abarth@webkit.org) The commit-queue is continuing to process your patch.
WebKit Commit Bot
Comment 6 2011-02-15 05:09:27 PST
Comment on attachment 82382 [details] Patch Clearing flags on attachment: 82382 Committed r78548: <http://trac.webkit.org/changeset/78548>
WebKit Commit Bot
Comment 7 2011-02-15 05:09:32 PST
All reviewed patches have been landed. Closing bug.
Gavin Peters
Comment 8 2011-02-15 05:51:08 PST
Renaming this bug (a bit late!) at ap's suggestion.
Simon Fraser (smfr)
Comment 9 2011-02-15 14:34:38 PST
Shame the changelog didn't say anything about the changes.
Note You need to log in before you can comment on or make changes to this bug.