Explain a little about how oed scaling operations are being applied.
This commit is contained in:
1 parent
634ea01ff0
commit
c24d00d8d4
1 file changed
+12
@@ -3030,6 +3030,18 @@ mged_escale(struct mged_state *s, fastf_t sfactor)
|
||||
|
||||
/* Have scaling take place with respect to keypoint,
|
||||
* NOT the view center.
|
||||
*
|
||||
* The changes written to model_changes are what are ultimately used by moveHobj
|
||||
* or moveHinstance to change the on-disk geometry. The job of updating what
|
||||
* is being shown in the MGED window for editing purposes is handled by the
|
||||
* new_edit_mats call, which updates view_state->vs_model2objview. dozoom.c
|
||||
* uses that matrix to modify how solids involved in editing are displayed.
|
||||
*
|
||||
* Note that this means wireframes for editing solids in oed mode aren't going
|
||||
* to be redrawn to reflect their new parameters. This is expected - it is quite
|
||||
* possible that the oed-distorted shape of a primitive will not be representable
|
||||
* using primitive parameters, so regeneration of new wireframes in response to
|
||||
* the new shape often isn't feasible.
|
||||
*/
|
||||
VMOVE(temp, s->s_edit->e_keypoint);
|
||||
MAT4X3PNT(pos_model, s->s_edit->model_changes, temp);
|
||||
|
||||
Reference in new issue
Block a user