On Mon, 2006-03-13 at 20:26 +0100, Juergen Spitzmueller wrote: > Martin Vermeer wrote: > > No it is not. The attached is (cursed evo interface!) > > Yes, that's better :-) > Works for me, from a quick test.
Do you think it is tested enough to go into 1.5? Attached the version with relevant comments added. - Martin
Index: changes.h
===================================================================
--- changes.h (revision 13312)
+++ changes.h (working copy)
@@ -84,7 +84,8 @@
/// return true if there is a deleted or unchanged range contained
bool isChangeEdited(lyx::pos_type start, lyx::pos_type end) const;
- /// remove the given entry
+ /// remove the given entry. This implies that a character was
+ /// deleted at pos, and will adjust all range bounds past it
void erase(lyx::pos_type pos);
/// output latex to mark a transition between two changetypes
@@ -134,22 +135,23 @@
typedef std::vector<ChangeRange> ChangeTable;
- /// our table of changes
+ /// our table of changes, every row a range and change descriptor
ChangeTable table_;
/// change type for an empty paragraph
Change::Type empty_type_;
- /// handle a delete
+ /// handle a delete, either logical or physical (see erase)
void del(Change change, ChangeTable::size_type pos);
- /// handle an add
+ /// handle an add, adjusting range bounds past it
void add(Change change, ChangeTable::size_type pos);
- /// merge neighbouring ranges
+ /// merge neighbouring ranges, assuming that they are abutting
+ /// (as done by set())
void merge();
- /// consistency check
+ /// consistency check, needed before merge()
void check() const;
};
Index: CutAndPaste.C
===================================================================
--- CutAndPaste.C (revision 13339)
+++ CutAndPaste.C (working copy)
@@ -234,7 +234,8 @@
pit = last_paste;
pos = pars[last_paste].size();
- // Maybe some pasting.
+ // Join (conditionally) last pasted paragraph with next one, i.e.,
+ // the tail of the spliced document paragraph
if (!empty && last_paste + 1 != pit_type(pars.size())) {
if (pars[last_paste + 1].hasSameLayout(pars[last_paste])) {
mergeParagraph(buffer.params(), pars, last_paste);
Index: paragraph_funcs.C
===================================================================
--- paragraph_funcs.C (revision 13339)
+++ paragraph_funcs.C (working copy)
@@ -236,6 +236,19 @@
pos_type pos_end = next.size() - 1;
pos_type pos_insert = par.size();
+ // What happens is the following. Later on, moveItem() will copy
+ // over characters from the next paragraph to be inserted into this
+ // position. Now, if the first char to be so copied is "red" (i.e.,
+ // marked deleted) and the paragraph break is marked "blue",
+ // insertChar will trigger (eventually, through record(), and see
+ // del() and erase() in changes.C) a "hard" character deletion.
+ // Which doesn't make sense of course at this pos, but the effect is
+ // to shorten the change range to which this para break belongs, by
+ // one. It will (should) remain "orphaned", having no CT info to it,
+ // and check() in changes.C will assert. Setting the para break
+ // forcibly to "black" prevents this scenario. -- MV 13.3.2006
+ par.setChange(par.size(), Change::UNCHANGED);
+
Change::Type cr = next.lookupChange(next.size());
// ok, now copy the paragraph
for (pos_type i = 0, j = 0; i <= pos_end; ++i) {
signature.asc
Description: This is a digitally signed message part
