Debugging window positioning under X11

[Stefan Monnier]

Could someone give me some hints where I should try and dig in order to
solve some window-positioning problems I have?

- seems to only appear with the athena/lucid toolkit, not with gtk.
- does not appear with metacity, but does appear with my venerable ctwm.
- appears both with the emacs-23 and the trunk code.  An old Emacs-22
  I have around does not exhibit this problem.

The problem:
- emacs -Q -g 80x40+0-0
  does not show up where I expect it: the frame is a lot lower
  than planned: I barely get to see the top half.  With gtk it's either
  a bit lower or a bit higher than planned, depending on whether
  I activate tool-bar-mode, but it's not nearly as wrong as with athena/lucid.
- with 80x30+0-0 I barely see the top of the frame.
- with 80x29+0-0 (or anything below 30), it gets placed at +0+0,
  probably because ctwm tells itself "wow, it's outside the screen,
  let's ignore the specified geometry".
- looking at the x-dimension position of the frame, it looks like the
  distance between the top and the frame's position is about twice what
  it should be (e.g. when the -0 corresponds to +197 the frame's
  placed at +450, when the -0 corresponds to +347, it ends up at
  +750 instead, and when -0 corresponds to +497 it ends up at +1034).
- after moving my minibuffer frame to where I like it to be, the frame
  params still say (top . 0) (left . 0).
- when I run emacsclient, which opens a new frame, the minibuffer frame
  jumps back to +0+0.

Does this ring a bell for someone?
Any hint where I should start digging?


        Stefan