That is a programming error if precondition is not fulfilled.

Change-Id: I9eb04cc99d10c27f193321ecaee74cc1b41ce531
Reviewed-on: https://gerrit.libreoffice.org/65359Reviewed-by: Eike Rathke <erack@redhat.com>
Tested-by: Jenkins

Move XIndexAccess Java tests to C++ for ScTableRowsObj.

Change-Id: I90dc592ba7901dab9c402b56e9e992fe7593c71c
Reviewed-on: https://gerrit.libreoffice.org/65371
Tested-by: Jenkins
Reviewed-by: Jens Carl <j.carl43@gmx.de>

Change-Id: I4d8fb95af78330d488e4b782eb8cdff5b83f1cc2
Reviewed-on: https://gerrit.libreoffice.org/65369
Tested-by: Jenkins
Reviewed-by: Jens Carl <j.carl43@gmx.de>

The new mode would consider a fly anchored at a node that contains the
start or end of a redline but has no visible text displayed to be
hidden, while the old move-code would move it to the merged SwTextNode
and therefore keep it visible.

Change this to give the same result both on layout creation
AppendObjs() and switching AddRemoveFlysAnchoredToFrameStartingAtNode().

Reproduces with ooo108073-2.odt, tdf104735-1.odt

Change-Id: Ibac795dcf7848d78e8a2200124bc2557f35ba434
Reviewed-on: https://gerrit.libreoffice.org/65366
Tested-by: Jenkins
Reviewed-by: Thorsten Behrens <Thorsten.Behrens@CIB.de>

Change-Id: I4bc3f2743b2d57308e687eb37f78c625c2bdc184
Reviewed-on: https://gerrit.libreoffice.org/65364
Tested-by: Jenkins
Reviewed-by: Jens Carl <j.carl43@gmx.de>

otherwise Desktop::Main will do it and soffice will crash w/ empty
error msg

Change-Id: I56768909227a6ff158ae353e1eff9b249d6ba9d4
Reviewed-on: https://gerrit.libreoffice.org/65357
Tested-by: Jenkins
Reviewed-by: Thorsten Behrens <Thorsten.Behrens@CIB.de>

not only executeQuery.

Change-Id: I2ee72078c48f622484dba46436f53990a45da364

note that we implement it in a very limited way, since we will always
return only a single result.

Change-Id: Idc7927d1a6896b78f6de8627ba857982821ac629

Change-Id: I3426df04b938487844e471c4cd34bcd72a1b3dfc
Reviewed-on: https://gerrit.libreoffice.org/64617
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>

don't need it anyway, the width req is already set by SvxNumberPreview::SetDrawingArea

Change-Id: Ic5b03a13a98c76c5ce927b13fc295510b6d8686d
Reviewed-on: https://gerrit.libreoffice.org/65353Reviewed-by: Xisco Faulí <xiscofauli@libreoffice.org>
Tested-by: Xisco Faulí <xiscofauli@libreoffice.org>
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>

Change-Id: I1bb4ee9354914db21d0440f83c156069a5efd9f9
Reviewed-on: https://gerrit.libreoffice.org/65355
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>

No need to suggest showing redlines beyond 250 changes anymore.

Change-Id: Ia219a4eb3af64079979538b8ab4aaf2a94ec19c8
Reviewed-on: https://gerrit.libreoffice.org/65294
Tested-by: Jenkins
Reviewed-by: Michael Stahl <Michael.Stahl@cib.de>

...able to be triggered by the UI, so the performance problem should be
gone.

The SwXTextDocument "ShowChanges" property allows API clients to call
SetRedlineFlags() and change the model; this API should be retained
as-is for now, because there is no way currently (outside of the rather
specialised a11y API) to get a "merged view" of the paragraphs in the
document, so API clients may want to (temporarily) toggle this property
to then retrieve properties from the model.  In case we later find that
there are no API clients that want to do such things, the property can
be removed.

A run with assert() instead of SAL_WARN_IF() found only 2 tests that
explicitly call SetRedlineFlags() to clear ShowDelete.

Change-Id: I44aee90f7727948ca0bbaaff9d04f8dd6b97cbd1
Reviewed-on: https://gerrit.libreoffice.org/65350
Tested-by: Jenkins
Reviewed-by: Michael Stahl <Michael.Stahl@cib.de>

Allow setting Qt5Object's position and size

Change-Id: I9f70b68ff402a3975e36baca4d81103603110d82
Reviewed-on: https://gerrit.libreoffice.org/65309Reviewed-by: Jan-Marek Glogowski <glogow@fbihome.de>
Tested-by: Jenkins

Move XNameAccess Java tests to C++ for ScScenariosObj.

Change-Id: I73dd65e672e1e98a6839cea03c30cfafb7da4ff4
Reviewed-on: https://gerrit.libreoffice.org/65313
Tested-by: Jenkins
Reviewed-by: Jens Carl <j.carl43@gmx.de>

Move XElementAccess Java tests to C++ for ScScenariosObj.

Change-Id: Ic1592746f99f685375e5b96c5b51c221d48ab07f
Reviewed-on: https://gerrit.libreoffice.org/65308
Tested-by: Jenkins
Reviewed-by: Jens Carl <j.carl43@gmx.de>

remove ExperimentalMode checks

Change-Id: Ie098eda0840bbd3231696cae43cc572ad61379fc

This assert is wrong, because there may be a legimate need to query
a position that is currently hidden.

However, oddly enough, we haven't found a legitimate document yet that
triggers this, and it has pointed us to several interesting functions
that needed adapting to merged frames... so keep it a little while and
remove it (and return something other than COMPLETE_STRING from mapping
functions) later...

Change-Id: Ide2abf0c2ce054b75c917063a5d5aa8d13430426

There's a test for that, testRedlineFlags() in sw_globalfilter

Change-Id: I9c821732910298d74271d95f0bfd0e902af84dbd

Change-Id: Ief227f3f5b03ec186178f4d68c3a415969d65507

nullpointer in ~XMLRedlineImportHelper in
SwUiWriterTest::testThreadedException()

Change-Id: Ic4a79ceffa831aced4db47836333a15a13773887

Change-Id: I78cfca1cd629bf518f8294450a34ee228a360ef9

* rename the only user, some UpdateFields() overload, to UpateOneField()
* restrict it to search for non-formatting hints; the FindAttrsImpl()
  should be used for formatting hints instead

Change-Id: I15002610a287fcdcd76777733a277d2ce64904bc

Iterate over merged nodes, and iterate over extents in them, so the
complex attribute-finding logic doesn't need to be changed.

Change-Id: Id72ca9247165a0c93f0cf2a38b7a0249771ca2eb

When redlining is enabled, the result will be a delete redline for the
existing text, and an insert redline for the new text; that much is
obious (and ReplaceRange can deal with a selection larger than one
SwTextNode easily, since it mostly adds redlines).

For the case when redlining is disabled, there are 2 options,
and i don't really know which is preferrable from UX point of view:

One approach is to reuse GetRanges() to ignore delete redlines in
the replace range; move its declaration to different header.

Another approach is to DeleteAndJoin() the existing delete redlines,
which is the same as the previous model based Hide mode,
which calls DeleteRedline() to remove the hidden redlines.

Also change ChgAutoCorrWord() to call DeleteSelImpl() directly.

Change-Id: I5974409d09eb39e04cc0b5dfc20d4db510e1cf58

Generally speaking, the find functions should find the strings that are
visible in the document view.  However, they may also be called from
SwXTextDocument functions, where they ought to find the strings that are
actually in the document model.

So concoct some funny helper types and adapt this so it can handle both
cases; it's not pretty but maybe it even works.

Change-Id: I1917398ff928e922673353e75e8fb724dc042031

Also, some of these don't really make much sense as member functions of
SwPaM, so make them plain functions instead.

Also, some pointer parameters are never null, but there is code to
handle that case in MakeRegion(); simplify that.

Change-Id: I822e814b0cab14f7f31cfa33478ed15774405a61

... so use it only once.  The problem is that the second time it will
call DelFrames() on every non-first node, which then empties the first
frame's newly created MergedPara instance.

This can be reproduced by toggling Edit->Tracked Changes->Show in
tdf99766-2.odt.

Also change some function parameters on e.g. SwTextFootnote::DelFrames()
to SwRootFrame to make it obvious that these don't actually need a
particular SwTextFrame.

Change-Id: I129be5db4a30546905c24c5e8820a5f105a33109

... for now, because this is the only thing that deletes existing
sw::MergedPara instances.

Change-Id: If092df52e63de28de38129f0aa11ad08156855c4

The nTFIndex is used to insert into the merged string; the correct index
is at the position corresponding to itInsert in the extents array.

Once the itInsert is found, the nTFIndex is no longer used inside the
loop, so just stop updating it.

Change-Id: Idc073cc758bad42b0d8b44d15f0f8e732eeb2a25

It turns out that there was a small problem in the interpretation of
sw_JoinText(), or rather, its caller,
SwRangeRedline::DelCopyOfSection(), which, since about OOo 3.2 and
i#100466, passes in bForceJoinNext, so the result is that the first node
wins always, not just for RES_BREAK/RES_PAGEDESC items.

This means that pParaPropsNode and pFirstNode are the same thing really.

Another little problem is that the SwAttrIter was initing the font
wrongly: the relevant items are the items in the *current* node's item
set on top of the item set of the paragraph style, i.e. the *first*
node's style.

Simple reproducer: ooo79457-1.odt

Change-Id: I06ef3c1695b8f3cdbded238864a60d5eb9ce4c44

The iteration was stopped as soon as the first start-node of a table
cell was reached, because the SwTableNode case didn't skip over the
table section in the nodes-array.  As seen in ooo67621-1.odt.

It looks like the same problem can't happen with sections or flys
because only tables have nested SwStartNodes and the Fly's nodes
can hardly all be in a delete redline.

Change-Id: Ib21a758c588d5dc8193ba171353f872de71bd65a

As seen in ooo50225-1.sxw.

Change-Id: I9f0bc0838cbf216b9a73e3ce6bc06b4340064077

SwCursorShell::GotoMark(), GoNextBookmark(), GoPrevBookmark() check
visibility of mark in layout.

SwWrtShell::GotoMark() returns whether it was successful.

EnhancedPDFExport skips marks that are not visible when creating links.

Change-Id: I36c0706c18a5f861b15eaf451a97daec04104056

SwEnhancedPDFExportHelper should not create links to things that are
hidden by the layout:

* SwCursorShell::GotoFormatField() skips deleted fields
  (this triggers an assert in GetCharRect() with ooo69593-1.odt)

* SwEditShell::GetINetAttrs() skips deleted INet attributes
  (which, as an additional bonus, hides then in the Navigator too)

* SwDoc::GotoOutline() skips outline nodes where we know they have
  been deleted (as seen in ooo66088-9.odt)

* SwFlyFrameFormats are skipped by EnhancedPDFExportHelper itself

* footnotes dito

Change-Id: Ife77dc7724688631d20cbaf8531b3c826a8ece94

moz55970-4.odt has a delete redline containing
 <text:p text:style-name="Standard">&#13;</text:p>
which the silly whitespace normalisation of course imports as an empty
SwTextNode.

A delete redline containing no text (or at least a fly?) appears quite
useless as the user can't do anything with it, so just delete it on import.

Change-Id: I657eeb627235adf90d7d7c1e44f505e2c6eacd46

It must use the start position that GetNextRedln() returned, not the
start of the end redline that GetNextRedln() returned, obviously.

This triggers an assert in CanSkipOverRedline() in ooo31961-27.odt.

Change-Id: I40b9647e0b8be9768e49047d309dd22470e658bd

This is really stupid...

Change-Id: I487ded0c0268ee9b7426e08109a9e8d447a0dfac

As seen in ooo95711-1.odt, the ODF import may create a redline that
starts on a SwTableNode - though i haven't been able to figure out how
such an odd creature might arise from UI actions.

Try to fix the obvious places so we don't crash easily; there might be
more trouble elsewhere though with code that assumes that a redline
starts on a SwTextNode.

Change-Id: I8431c1416ac4503ff0209a946398656f1c28366d

This was another IsDelLastPara() redline in ooo27109-1.sxw.

Change-Id: I1f4198a4bcbed7bdd52991af3cb3322ebacc65fe