It would certainly be immediately obvious to any reader of the RTF spec
that \sN will apply the style with index N to the current paragraph.
But actually, that is not what Word does when it reads \sN...
what it really does is to apply the style with index N, and then for
every attribute in that style, apply the same attribute with a default
value to the paragraph, effectively overriding what's in the style.
If that doesn't make any sense to you, well, have you heard the joke
about how many Microsoft engineers it takes to change a light bulb?
Also, \pard apparently implies \s0.
To implement that, change RTFSprms::deduplicate() to recursively
look for style SPRMs that are missing in the properties, and put
in default ones, currently just for 2 keywords \sa and \sb.
This requires changing deduplicate() to be const and return a new value,
since it is no longer idempotent, as the erased SPRMs would get
defaulted on the next run.
While at it, fix RTFValue::equals() which did not compare m_sValue.
This fixes the testParaBottomMargin test that was broken by the fix
for fdo#70578.
Change-Id: I4ced38628d76f6c41b488d608a804883493ff00b
quoth the spec:
"For <style>, both <styledef> and <stylename> are optional; the default
is paragraph style 0."
Of course in order to do that we need to add support for at least
recognizing the \dsN and \tsN keywords to override the default, so that
table styles don't become paragraph styles.
Change-Id: Ic100768581f9e8c327063ff776fbd61ac4242483
RTFSdrImport::resolve() is called for \picprop and creates an XShape
that is stored in RTFSdrImport::m_xShape and also
DomainMapper_Impl::m_aPendingShapes;
later RTFDocumentImpl::resolvePict() completely ignores that XShape
and creates a new one, which is also inserted in the document;
the first XShape is effectively leaked.
Try to avoid that by re-using the exising m_xShape in resolvePict().
Not sure if there are any problems with doing this, it's all a bit
confusing.
Change-Id: I98456242acb0766f547eb8f7d877f51d53323f3a
If the shape properties are inside \picprop destination, don't set
shapeType.
(regression from 9f1f7199736e2ae07b34849ba66f61a1ef5782e8)
Actually this does not fix the root cause, this is just a work-around,
the extra shape is still inserted but it's invisible now.
Change-Id: I6cf093de2a5657533f393863ed8010ae083bec16
<vmiklos> sberg: the comment says that it was 2, so unless you're interested in
debugging it, i would just lower the barrier (423 -> 369)
Change-Id: I0269eab4eee594c4d7aad1c710ed389443208bc4
This is a variable-length encoding, and the second byte may be a RTF
syntax character like \, {, }.
Change-Id: I813ccafda18388af3bf05eb7ce9a0253c627b1c4
If the old width was that small, then something larger should be also
fine. (Seen 564 instead of 423.)
Change-Id: Ife0fd4a55738268cdc1691b90e28b51adba829ef
In the past, LO did not support 'outset' and 'inset' border types
for tables, so when encountering them - LO converted them to other
types of border. Now that LO supports it - DOCX and RTF filters
were changed to import these border types correctly.
DOC filter was not changed, because creating a proper DOC sample
file that has these border types (needed for the 'bordertest.hxx'
file) was not possible right now. So at least DOCX and RTF filters
are fixed.
Change-Id: Ida2449d45a0ac138388f3cbfeb41657db1d4cda9
Signed-off-by: Michael Stahl <mstahl@redhat.com>
Previously only the number of nested sprms / attributes was compared.
With this, the font of the bugdoc is correctly Arial, not Times.
Change-Id: I351de414b6734336b31c1334dbd2349072f16002
Regression from a48e2fd9049797110b3b2505c363557284987ca8 (fdo#44736 -
convert RTFSprms to a copy-on-write structure., 2012-12-07)
Change-Id: I2538f440e29cef6d40db2ea914e4afcbfe411890
The Writer 'Heading 3' paragraph style is gray by default, but (just
like in case of DOCX) that shouldn't have any influence on the RTF
import result.
Fix this by moving the compat setting from the DOCX filter
implementation to the common dmapper.
Change-Id: I86c7cf1a66f82b438ce8379467773a88c9e229af
This scenario is not a valid one, Word doesn't handle it, either -- but
the old LO 3.4 parser did. So add minimal support for it to avoid
the regression.
Change-Id: Icc2e8d3bf314e9cadda57956668033aa6d092457
Commit bbe3627eece0c3486e7ea11f2f13377aaa3a8fed (rtftok: stop sending
sprm:CRgFtc{0,1,2} tokens, 2014-03-05) dropped support for case when a
font name is used in multiple entries in the font table, but with
different encodings.
Turns out that this is a valid use-case, so revert back to the old
behavior where the key of the encoding table is the font index, not the
font name.
Change-Id: I048dff58af801d704fd4bc75a6a4dcb0f03bf185
Commit 6e61ecd09679a66060f932835622821d39e92f01 (Merge back branch
alg_writerframes to trunk, 2014-03-19) changed the sw UNO API, so that
in case SvxBrushItem / RES_BACKGROUND was to be get/set, it mapped that
to svx's fill style.
This has two problems: 1) it does not work (see the failing unit test)
and also page style editing UI sets SvxBrushItem, so even if the UI is
used to (actually successfully) set the page background, it won't be
saved to ODF.
So, do this background <-> fill style mapping only for frames for now.
Change-Id: I734880a40c891ae9e341cbddf8f88b0462361aac
Commit 1eaab77c718ffa254068ae6032862dfb5a03db67 (fdo#60722 import
RTF_SHPZ, 2013-03-06) changed how we handle z-order, in case two shapes
have the same value. Turns out for drawing-objects the order is the
opposite in this situation.
So fix this by adding a new mode, that keeps the original testcase happy
without breaking older documents.
Change-Id: Ib2d284cefc3c0dce40ac2e516ba260d6cd04ce43
The problem was that dobxpage arrived first, set HoriOrientRelation to
FRAME, then dptxbx tried to apply defaults, which overwrote the already
set HoriOrientRelation. Fix this by only applying properties which are
not set yet.
Change-Id: I108f3363a2758eee0242533fe92e511e8c522b68
When a document ends with \sect it's possible that a spurious page break
is created. In fact the spurious page break is always created by the
RTF importer, sometimes it is deleted again by
DomainMapper_Impl::RemoveLastParagraph() and sometimes not.
It is created because on the final \sect RTFDocumentImpl::sectBreak()
still calls startSectionGroup(), and the popState() for the \rtf1 group
then calls sectBreak() another time.
To prevent this, do not call startSectionGroup() from sectBreak() but
instead from setNeedSect(), and ensure that it is called as soon as
anything after \sect is read.
One unit test fails because the \page is not handled properly: the
conversion to \skbpage \sect \skbnone is not correct, because the \skb*
keywords are an exception and affect the \sect that precedes them, not
the following one; sending the \skbpage later unfortunately requires
additional cleanup later.
Change-Id: I3c1a3bceb2c8b75bbecdc748170562451ce5f5c3
The previous fix for this bug only fixed a symptom, this a fix for the
real problem; with the real problem fixed the nCellEnds is unnecessary.
Given that top-level table properties may be put either before or after the
table cells, the only way that works to import tables is to buffer a whole
top-level table row, but currently the buffer is replayed already at the
end of a nested table row.
Fortunately the RTF spec guarantees that \nesttableprops must occur
after the nested table cells of the nested row, so it should be
sufficient to remember the cell properties for the current nested table
row only, in addition to the cell properties for the top-level table row.
With this change, skipping a \nesttableprops destination when there is
a table style turns out to mangle ooo98040-1.rtf badly, so stop doing
that workaround.
RTFDocumentImpl::popState() was copying various buffers up the state
stack which is a clear indication that these shouldn't be members of
RTFParserState in the first place, move them to RTFDocumentImpl.
Change-Id: Ic2d8f7b3e00844b224d61605b405ca651239e5f7
Added vcl/settings.hxx to all cxx files which require it.
This helps to speed up compilation after changes to the settings.
Conflicts:
sc/source/ui/dbgui/pvlaydlg.cxx
Change-Id: I211a0735c47f72d6879f6f15339355abfe0e3cf4
Reviewed-on: https://gerrit.libreoffice.org/7933
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
In this document written by "XMLmind XSL-FO Converter" there are less
\cellx than \cell and thus when reading \nestrow/\row a whole buffered
nested table \cell is lost and then subsequently the rest of the nested
table too. Try to fix that by counting both \cell and \cellx and
replaying until the maximum of those.
Cannot count \intbl since we synthesize that in various places.
(regression in LO 3.5)
Change-Id: I3b64ad94af842e076611418589a0c83bd18841c6
Commit 330b860205c7ba69dd6603f65324d0f89ad9cd5f (fdo#68787 DOCX import:
handle when w:separator is missing for footnotes, 2013-09-04) disabled
footnote separator by default in dmapper, as the OOXML tokenizer always
provided a uFtnEdnSep in case a separator was wanted.
Let the RTF tokenizer do the same, this way we're in sync with Word
again: if RTF_CHFTNSEP is in RTF_FTNSEP, then we show the separator,
otherwise we don't.
Change-Id: I74b46c5d71227682e093695336dc9eb6fde22121
And not the other way around, how
24ee3df385cf2aa95cd888581c84fdf90cc682dc (RTF import: fix priority
handling of shpz vs dhgt, 2012-04-10) did, this time with a reproducer.
Change-Id: I9412341c6b35ca2760e4490a18f11bc6a0e0b78a
