Bug sf-1741659: annoying decimals in coordinates

[Ineiev]

This misfix adjusts the labels to grid when applicable.

There is two other ways, reduce the precision output
and process the grid in floating point; the former
makes the labels less useful, the latter would produce
more much more complicated patch.

[Felix Ruoff (felixruoff)]

The patch of Ineiev here seems to be newer than the one in LP:699169. It applies and runs without problems on my system, like the patch in the other thread, too. So, I would very welcome, if one of these patches got applied to HEAD soon :-) (The patch will close both bugreports, I think)

[Peter Clifton (pcjc2)]

Why does the patch have this change?

- ghid_cursor_position_relative_label_set_text ("r __.__; phi __._; __.__ __.__");
+ ghid_cursor_position_relative_label_set_text ("__.__ __._; __.__ __.__");

Other than that, it looks ok to me.

[Ineiev (ineiev)]

It is not related to the bug; I thought these letters are not really needed
(in the Lesstif HID they are absent); if you think they are, I'll exclude this change.

[Peter Clifton (pcjc2)]

Leave the letters as they were.. they add clarity to the meaning of the numbers.

I look forward to the fix.

[Ineiev (ineiev)]

Ok, I'll leave them; however, in Sept 2009 I discovered a regression in this patch;
I fixed it in my branch on repo.or.cz, but I need to check all again before pushing.

I think I'll do it within a week.

[Ineiev (ineiev)]

Now, the minimal fix.

The minor problem is loss of precision
in many places (where unprecise constants
for metric-to-imperial conversions and floats
for crosshair calculations are used).

The major problem is absence of any reliable means
to distiguish between metric and non-metric grids
(since Settings.grid_units_mm is really an attribute
of display and doesn't define whether the grid is metric).

[Ineiev (ineiev)]

This is a testcase (a large board with two pads; the first
one is misaligned with metric grids by one PCB unit,
the second one is misaligned by two PCB units).

[Felix Ruoff (felixruoff)]

The patch looks fine to me.

But... I have a Problem with the testcase-file: I can select objects/elements/lines, but I can't deselect them! I have also tested this with Version 20091103 (lesstiff GUI and GTK+ GUI) and there is the same Problem (so I think, this Problem is not related to this Patch, but perhaps someone has an idea to fix this!),

[Ineiev (ineiev)]

Thanks, pushed!

WRT the text selection, it is an unrelated bug: the bounding box is not assigned
correctly for some reason. I was able to "drag" the texts from the centers of the pads
(both before and after applying the patch).

[Stephen Ecob (silicon-on-inspiration)]

Thank you, the patch makes my life better :)

I notice that it works fine for the absolute coordinates, but the relative coordinates are sometimes still broken - see the attached screenshot. The problem seems to only come up for some positions of the reference mark (set with ctrl-m).
The screenshot was taken with units set to mm and grid set to 0.1mm
PCB built from git head as at [Tue, 22 Mar 2011 20:43:30 +0000]

[Ineiev (ineiev)]

Thank you, should work now.