commit 9f01951b85 also did the extra
Remove/Add for Draw fly objects, and it turns out that that's actually
wrong because SwTextFlyCnt::SetAnchor() will set the anchor without
locking anything if it's a Draw object. Replace it with a different
hack in SetAnchor() that applies only if it calls LockModify().
Thanks to Varun Dhall for creating a reproducer document.
Not sure if the LockModify() could be replaced completely, perhaps it's
just an optimization to avoid re-creating layout frames for the fly.
Change-Id: Ib3236f289c2c4202d48ac378a53ce02130d4ce2c
Start Writer, Insert Image, Anchor as character, Go after
image, press enter, writer crash
This is because m_pAnchoredFly is not updated.
JoinPrev, JoinNext and SplitContentNode all rely on CutText with calls
InsertHint. InsertHint calls SetAnchor. SetAnchor calls Modify callback
except if "LockModify"ed. This patch ensures that, whatever the value of
LockModify, the references are kept correct.
Change-Id: Id7254784c6954db4b542b2c4228b388fb924bbc2
Reviewed-on: https://gerrit.libreoffice.org/16041
Reviewed-by: Michael Stahl <mstahl@redhat.com>
Tested-by: Michael Stahl <mstahl@redhat.com>
The old internal RTF filter used to call SwTxtNode::SetAttr() without
setting SetAttrMode::NOFORMATATTR, so character attributes which cover
the whole node got converted to paragraph attributes. The new UNO
filter goes through SwXText::insertTextPortion(), which sets
SetAttrMode::NOFORMATATTR, so this doesn't happen. The result of this is
that when the UI sets a new paragraph style on the text node, then such
character attributes are no longer removed.
Given that in RTF you can't really have character properties on a
paragraph, going back to the document model produced by the old internal
filter doesn't sound like the good direction -- not to mention that
changing SwXText::insertTextPortion() this way would be an implicit API
change.
Fix the problem by tweaking SwEditShell::SetTxtFmtColl() instead, so
that it removes these full-text-node character attributes, too. The
logic in SwTxtNode::RstTxtAttr() can be extended later if necessary to
delete more attributes, but to be on the safe side, just handle the bare
minimum necessary to fix the problem for now.
Change-Id: I5faa3226fc0f9c18e005da185fe0830d8c393256
SwRangeRedline::Show() will move nodes around in the nodes-array, which
means that using SwNodeIndex as a key in a map that has a lifetime
not limited by the stack is a bad idea, as the map will become unsorted.
Remove SwFrmFmtAnchorMap from SwDoc and replace it with new
SwNode::m_pAnchoredFlys to do the same mapping.
(regression from 738fb2ad77)
Change-Id: I396d92b9d0b2045e98bad6d0b374303cd4e62b59
Turn the Link class into a template abstracting over the link's argument and
return types, but provide default template arguments that keep the generic,
unsafe "void* in, sal_IntPtr out" behvior. That way, individual uses of the
Link class can be updated over time.
All the related macros are duplicated with ..._TYPED counterparts, that
additionally take the RetType (except for LINK_TYPED, which manages to infer the
relevant types from the supplied Member).
(It would have been attractive to change the "untyped" LinkStubs from taking a
void* to a properly typed ArgType parameter, too, but that would cause
-fsanitize=function to flag uses of "untyped" Link::Call.)
Change-Id: I3b0140378bad99abbf240140ebb4a46a05d2d2f8
Move all logic out of destructors, so it cannot happen any more that
members of a sub-class are accessed from a superclass destructor, when
those members are already dead.
Logic is now in virtual DestroyImpl() methods.
All SwFrms must be deleted with SwFrm::DestroySwFrm().
Change-Id: Icec5b12e12d5a2d955cb5844d7d4f7ac85ab79cd
since their usage is intertwined.
Also introduce new o3tl utilities enumrange and enumarray to make
working with scoped enums a little simpler.
Change-Id: I2e1cc65dd7c638e59f17d96dfae504747cad6533
When a SwTxtFormatInfo is created to format a paragraph, pre-compute the
result of vcl::ScriptRun::next() and cache it for future calls to
OutputDevice::GetTextBreak() and GetTextWidth().
This requires adapting a bunch of methods to pass the additional
parameter, and some classes to backup and restore the cache when they
replace the text of the SwTxtFormatInfo.
There is some code in vcl OutputDevice::ImplPrepareLayoutArgs()
to modify the passed string and replace digits depending on
"meTextLanguage" member; try to set it to the correct value when
creating the layout cache (unfortunately it's not possible if the user
sets the CTL Numerals config to the non-default "Context" value).
Another issue is the check in OutputDevice::ImplLayout() if there is
a mpConversion member on the font; apparently this is used to translate
between different Symbol fonts, so not very important; just ignore the
cache in this case.
This reduces vcl::ScriptRun::next() from 11 to 0.36 billion callgrind
cycles when built with GCC 4.9.2 -m32 -Os (which is still 16% of the
formatting).
Change-Id: I61fb8530333f2e7a9199f767c00cf2181ba49951
Reviewed-on: https://gerrit.libreoffice.org/14732
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
When a SwTxtFormatInfo is created to format a paragraph, pre-compute the
result of vcl::ScriptRun::next() and cache it for future calls to
OutputDevice::GetTextBreak() and GetTextWidth().
This requires adapting a bunch of methods to pass the additional
parameter, and some classes to backup and restore the cache when they
replace the text of the SwTxtFormatInfo.
There is some code in vcl OutputDevice::ImplPrepareLayoutArgs()
to modify the passed string and replace digits depending on
"meTextLanguage" member; try to set it to the correct value when
creating the layout cache (unfortunately it's not possible if the user
sets the CTL Numerals config to the non-default "Context" value).
Another issue is the check in OutputDevice::ImplLayout() if there is
a mpConversion member on the font; apparently this is used to translate
between different Symbol fonts, so not very important; just ignore the
cache in this case.
This reduces vcl::ScriptRun::next() from 11 to 0.36 billion callgrind
cycles when built with GCC 4.9.2 -m32 -Os (which is still 16% of the
formatting).
Change-Id: I61fb8530333f2e7a9199f767c00cf2181ba49951
Reviewed-on: https://gerrit.libreoffice.org/14732
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
regression from
commit 553a058ec8
Date: Sat Mar 14 20:43:43 2015 +0100
core: dont use 'shortcut' that is longer than the original
Change-Id: I0154e484a4f4d6ce6672ddf35d1869157a61f71e
Change-Id: I4159a5caf492900bcfb863bc0bc5cf59a47a311c
