624 bytes, application/xhtml+xml
460 bytes, application/xhtml+xml
2.68 KB, patch
|Details | Diff | Splinter Review|
Created attachment 352264 [details] testcase ###!!! ASSERTION: the pointer to this sibling will be overwritten: '!aNewFrame->GetNextSibling()', file /Users/jruderman/central/layout/generic/nsFrameList.cpp, line 176 ###!!! ASSERTION: Some objects allocated with AllocateFrame were not freed: 'mFrameCount == 0', file /Users/jruderman/central/layout/base/nsPresShell.cpp, line 675 Likely exploitable. The testcase doesn't cause a crash, but a variant (created with the usual trick) makes Firefox dereference 0xdadadaf6.
The usual suspects, the block reflow reports incompleteness, the table starts to split, but as it is not in pagination mode, this goes wrong.
the crash is preceeded by: ###!!! ASSERTION: Shouldn't be incomplete if availableHeight is UNCONSTRAINED.: 'aReflowState.availableHeight != NS_UNCONSTRAINEDSIZE', file d:/moz_src/src/layo ut/generic/nsBlockFrame.cpp, line 1413
Should be fairly easy to fix once we figure out where the incompleteness is being triggered.
we trigger it by asking to reisze to a very small width of the column frame ColumnSet(div)(0) 05CCE828 Reflow a=60,UC c=60,UC h-resize v-resize cnt=10999 block 05CCE774 Reflow a=60,1200 c=60,UC dirty v-resize nif=05CCF8A4 cnt=11000 text 05CCE9F0 Reflow a=60,UC c=UC,UC dirty v-resize nif=05CCF838 cnt=11001 text 05CCE9F0 Reflow d=600,1140 status=0x1 text 05CCF838 Reflow a=60,UC c=UC,UC dirty v-resize pif=05CCE9F0 cnt=11002 text 05CCF838 Reflow d=660,1140 block 05CCE774 Reflow d=60,1200 status=0x3o=(0,0) 600 x 1200sto=(0,0) 600 x 1200 ColumnSet(div)(0) 05CCE828 Reflow d=60,1200 status=0x3o=(0,0) 600 x 1200sto=(0,0) 600 x 1200 ###!!! ASSERTION: Shouldn't be incomplete if availableHeight is UNCONSTRAINED.:'aReflowState.availableHeight != NS_UNCONSTRAINEDSIZE', file d:/moz_src/src/layout/generic/nsBlockFrame.cpp, line 1413
FWIW: I hit the assertions from comment 0 and comment 2 on the first testcase, and the incomplete assert on the second testcase, using my linux mozilla-central debug build (updated yesterday). Platform --> ALL/ALL
FWIW, the testcase on bug 435664 triggers this bug's assertion ("the pointer to this sibling will be overwritten") when printed / print-previewed. (It also crashes, which is what that bug is primarily about.) See bug 435664 comment 13. Possibly related to this bug?
I think the problem here is that the column set is propagating incompleteness that's ok inside of it to outside of it... although maybe incompleteness isn't ok inside the last column of a column set.
Created attachment 361492 [details] [diff] [review] fix The comment should help explain what's going on. Basically we're just changing the number of columns from 2 to 1. This should mean that the first column pulls all the content out of the second column, but that doesn't happen because we skip reflowing the first column. We need to tweak the incremental reflow test so that we always reflow the last column if we're giving it unbounded height; it will want to pull content from its next-sibling.
Comment on attachment 361492 [details] [diff] [review] fix r+sr=dbaron
This fixes bug 476579 too.
Pushed http://hg.mozilla.org/mozilla-central/rev/c0dbc2a40bb1. I withheld the testcase.
Comment on attachment 361492 [details] [diff] [review] fix a1.9.1=dbaron
Although the assertion doesn't occur on the 1.9.0 branch the code being patched looks the same. Do we need this fix on the 1.9.0 branch as well?
Blocking 1.9.0 on the assumption the answer to comment 16 is "yes, we do".
Code-freeze for 18.104.22.168 is next Tuesday. If this patch works as-is please request approval, else we need a back-port patch. Thanks.
Comment on attachment 361492 [details] [diff] [review] fix Patch applies as-is.
Comment on attachment 361492 [details] [diff] [review] fix Approved for 22.214.171.124, a=dveditz for release-drivers
Checked into 1.9.0.
Neither of the two testcases trigger the assert in my debug 126.96.36.199 build on OS X. Have we seen these asserts on the 1.9.0 branch?
Crashtest added: http://hg.mozilla.org/mozilla-central/rev/d3e7c18f58fc