On 2009-08-28, Ole Jørgen Brønner olejorg...@yahoo.no wrote:
No, you're right, they don't.
If someone wants to work on this, here's what I after
a very quick glance over the code (that it's been ages
since I last looked into), think might be the plan:
- In input_do_refit, we have to rqgeom instead of window_do_fitrep.
The crux is that with FULL_BOUNDS we know the available space.
With the size policy set to something else, we only know the
area the managing WMPlex (or other region) is willing to give us
per the size policy. So we have to ask for as much as we'd like
to have with rqgeom... but to calculate the wanted height for our
requests, we have to assume that the width stays fixed.
Thus, input_calc_size has to be modified to support calculations
with infinite max_geom.h (virtually; INT_MAX or so).
- We can no longer use the modified input_refit in input_fitrep
(that gets alled by the managing object to do the resizing),
instead an updated version of the current code has to be used
there. (It's just a couple of lines.)
- It should be possible to get rid of the last_fp field, modifying
all code involving it, as we no longer make requests using the
last known available space.
- Of course, for nicer top-aligned queries, the completions
should expand downwards, and this demands a bit of extra work,
figuring out whether we're actually attached to a WMPlex,
and asking for the effective size policy to see if it is a
top-aligned one.
Probably I'm missing something, but that should get anyone
interested in this started. I have better things to do...
actually haven't been able to use Ion lately, because Cygwin
X has been in a crash every 5 minutes state for nearly a
couple of months now. Typical FOSScrap.
--
Off the dope since 2009;
http://iki.fi/tuomov/b/archives/2009/07/21/T17_26_09/
