What's happening here is that we enter an infinite loop by oscillating
between two states. The code assumes that (a) the available space will
never grow, only stay the same or shrink, and (b) that we should break
out of the loop if it stays the same. This also means we hit the
assertion about the available space growing every other time through the
loop.
This is in the inner loop in nsBlockFrame::ReflowBlockFrame that was
introduced in https://hg.mozilla.org/mozilla-central/rev/80ef9bb2c2e9 .
The problem is fundamentally a logic error in that code. The makes the
assumption that if you reduce the width available to a block formatting
context or replaced block-level element, its height does not shrink.
(The "replaced block" (really block formatting context) in this case, as
in the original testcase, is a scroll frame. I didn't debug the
original testcase enough to figure out what caused its sizing
characteristics, although a percentage-width image does seem like the
most likely candidate.)
Without the patch, the reftest test (but not reference) hangs, as does
the semi-simplified test in the bug (given a narrow window).
With the patch, neither the semi-simplified test in the bug nor the
reference hangs, and the reftest passes.
--HG--
extra : commitid : APy8PfXlvvz
Note that nsMathMLContainerFrame and its subclasses are unchanged since
they are not target of fullscreen (and thus no backdrop frame), and they
have an assertion to ensure we really don't pass any unexpected list in.
--HG--
extra : source : a1f7ff18a69cc116120de33f14ae62f576a4b55a
AutoTextRun now only needs a DrawTarget instead of an nsRenderingContext, and
similar nsRenderingContext/gfxContext-to-DrawTarget replacements can be
propagated a long way up the call graph. This patch replaces 93 occurrences of
nsRenderingContext and 135 occurrences of gfxContext with DrawTarget; that's
13% of them.
The patch is mostly plumbing changes. A couple of not-entirely-plumbing
changes:
- It adds a comment about the null check in
gfxGlyphExtents::GetTightGlyphExtentsAppUnits().
- A couple of functions simply had an unused gfxContext or nsRenderingContext
parameter removed, e.g. SetLineBreaks().
--HG--
extra : rebase_source : 8f56994bb4d254a86788b17ab2864ebc758a7e6b
The bulk of this commit was generated with a script, executed at the top
level of a typical source code checkout. The only non-machine-generated
part was modifying MFBT's moz.build to reflect the new naming.
CLOSED TREE makes big refactorings like this a piece of cake.
# The main substitution.
find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \
xargs perl -p -i -e '
s/nsRefPtr\.h/RefPtr\.h/g; # handle includes
s/nsRefPtr ?</RefPtr</g; # handle declarations and variables
'
# Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h.
perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h
# Handle nsRefPtr.h itself, a couple places that define constructors
# from nsRefPtr, and code generators specially. We do this here, rather
# than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename
# things like nsRefPtrHashtable.
perl -p -i -e 's/nsRefPtr/RefPtr/g' \
mfbt/nsRefPtr.h \
xpcom/glue/nsCOMPtr.h \
xpcom/base/OwningNonNull.h \
ipc/ipdl/ipdl/lower.py \
ipc/ipdl/ipdl/builtin.py \
dom/bindings/Codegen.py \
python/lldbutils/lldbutils/utils.py
# In our indiscriminate substitution above, we renamed
# nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up.
find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \
xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g'
if [ -d .git ]; then
git mv mfbt/nsRefPtr.h mfbt/RefPtr.h
else
hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h
fi
--HG--
rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
This was introduced in bug 209694 (gecko-dev 13a65028) but became unused
through bug 292295 (gecko-dev 4593df2a57) and bug 300030 (gecko-dev
31f18988).
--HG--
extra : commitid : 1YUuWOKeInm
This adds an additional retry loop in block reflow that we can only
trigger when reflowing a block formatting context (replacedBlock
non-null). It can retry in two different ways, either with a narrower
width but at the same vertical position (when
ReplacedBlockFitsInAvailSpace is still true) or at a new vertical
position (which is treated as a form of clearance).
Fortunately we don't have to worry about margins collapsing *through*
such a boundary since we're dealing with a new block formatting context.
Note that Chromium passes all of the new bfc-displace-* tests, although
it moves the block formating context down unnecessarily in
bfc-shrink-1.html (which we do neither before nor after the patch),
though agrees with the width we have after the patch (but not before the
patch).
--HG--
extra : commitid : FgDc4hjTxHp
Using the wrong begin() means that RFindLineContaining would hang
in the case the frame isn't on the overflow list. We should always
find it there though, if it wasn't found on the principal list, so
the bug is harmless apart from the assertion.
