>>>>> "Angus" == Angus Leeming <[EMAIL PROTECTED]> writes:
Angus> On Thursday 14 February 2002 10:25 am, Jean-Marc Lasgouttes Angus> wrote: >> >>>>> "Angus" == Angus Leeming <[EMAIL PROTECTED]> writes: >> Angus> Now repeatedly type and erase the letter "a" in front of the Angus> inset. >> Angus> Here I get a new graphics inset created with each key stroke. Angus> This may be something to do with my hacking here, but I don't Angus> think so since the new inset is created through a clone command Angus> which is in the core somewhere. >> I think this is undo creating and deleting copies of your >> paragraph. To be sure, set a break in gdb at the place where your >> graphics inset is created and look at the backtrace there. Angus> Well, that's the point. It's creating them but not deleting Angus> them. The counter patch I posted has a decrement in the Angus> insetgraphics d-tor. OK, I see what you mean. Yes, undo is currently leaking memory like a thief. There is code in undo_funcs.C to properly delete unused paragraphs (the DELETE_UNUSED_PARAGRAPHS define), but it is not turned on, because it needs testing. And indeed, Michael's purify runs showed that redo has problems. JMarc