On Wed, May 30, 2007 at 11:36:13PM +0200, Stefan Schimanski wrote: > > Am 30.05.2007 um 23:29 schrieb Jean-Marc Lasgouttes: > > >>>>>>"Dov" == Dov Feldstern <[EMAIL PROTECTED]> writes: > > > >Dov> Please, could someone respond regarding this patch from Elazar? > >Dov> It fixes issue (3) of > >Dov> http://bugzilla.lyx.org/show_bug.cgi?id=3551, which is a > >Dov> regression. > > > >Dov> It works. The concept makes sense. The implementation -- well, we > >Dov> don't love it, but we haven't gotten any better suggestions. (The > >Dov> question is: how to identify the difference between being "at an > >Dov> inset" vs. "inside it".) It's safe --- at the very worst, it will > >Dov> only affect bidi users. > > > >I still have the nasty feeling that the correction could be done in > >somewhere like coordOffset (bufferview_funcs.cpp), but since I do not > >have the energy to actually research it, I think indeed that the patch > >could be committed. > > > >Actually, coordOffset is not good because it also handles non-text > >insets. > > I will make a clean patch tomorrow morning. It adds another return > value to cursorPos which tell whether insets in deeper slices will be > put to the left or right of the returned position. This will require > of course a change in the signature of all the cursorPos > implementations is needed. But that's trivial. coordOffset will use > that then to compute the right cursor position. I think that that's > the right way to implement it. Accessing the bufferview cursor is > even wrong, e.g. for selection spanning from an RTL to an LTR > paragraph (look at my other patch for a working RTL selection > implementation). Hmmm, this feels like a proper solution, not a hack...
- Martin
