>> Pavel Sanda wrote: >> > i guess its too late for 5211? >> >> We don't have a patch, do we? > >i have already some work in my tree, but it has problems with master/child in >trunk and most notably it violates string freeze, Nothing is really broken.... Please think along how to solve it. It's by design that children documents are 'released' when a master is 'released'. This happens in the Buffer::dtor. Think of a situation in which a child document is included in multiple master. Closing one of the masters will automatically release this child. This is prevented in BufferList::releaseChild IIRC. This function checks whether there are other parents of this file. Now, we need the same sort of thing for reloadBuffer. I can think of a few scenarios: - we can move the logic of releasing children from Buffer::dtor to some higher level, - or we can mark this buffer in BufferList somehow to be in a 'transilient' state. So, BufferList::releaseChild knows that although there is no other master for this file, this master will be back again so the child don't need to be closed. >which came out of the blue for me. It sounds like this kind of 'new feature' will need some cross-platform testing, right ?
pavel
