On Tue, Oct 26, 2004 at 09:03:23AM +0300, Martin Vermeer wrote: > To add my spicing to this soup, shouldn't we first take inventory of > what bugs we actually have, and how critical they are? There is all > kinds of stuff on bugzilla, and I suspect some of them must be > absolutely fixed while others can be plastered over safely like Andre > suggests... but could "somebody" make a clear compilation on this?
There is at least one fat bug in undo.C when creating an item on the redo stack. The situation when everything stays within the same inset works fine, but e.g. C-n C-l C-z C-S-z crashes. This needs a real fix (basically figuring out which region of the doc is really affected, i.e. probably just going up one slice in the stored cursor and use this paragraph) Then there seems to be a problem with positioning cursor 'in the middle of nowhere' when e.g. using the scrollbar. This bug might have been there for a long time but was concealed by 'unnecessary' rebreak that kept the position caches at least hand warm. Now that we don't rebreak too often, caches are sometimes non-existent or completely out of sync and bite us as they are not just used as caches but also serve as 'primary data' for all cursor up/down related actions. This is a design problem. One obvious solution would be to fix the design, but that's out of scope for 1.4, the less intrusive solution is to keep the position caches sufficiently warm, i.e. to rebreak paragraphs as we scroll and/or similar drastic measures. These are the 'big' things I am aware of. Andre'
