Add circular grid to ronchi and Foucault images.

[githubdoe]

AI and I have been busy.

[githubdoe]

Added this functionality to the right click menu of ronchi image.

[githubdoe]

When I tried to make a pull request for this. It does not show up as a selection. I thought I knew how to do that. But...

[atsju]

Hi Dale,
yes it did not show up because you worked on same branch.

As always, start from master. One feature = one branch = one PR
When you start something new, you should ideally always checkout master, pull, and create a fresh branch.

Now we have 2 changes is same PR #331.
The last commit with these additions will require some changes because of newly introduced build warnings (see "file changes" tab).
The original commit were approved for merge.

@gr5 I propose to do a new branch and PR with original changes from Dale. This then can be merged and we can use the inital PR, rename it and use it for the new feature described here (just cose/reopen after the new one is merged to update scope). OR Dale to create a new branch but this will probably be too complex.

[gr5]

We hadn't merged the previous PR. So you could have merged that one and then pushed your new changes.

But instead your updates went into the existing, already approved PR yet not merged PR.

Once the PR is merged and closed then you can push new edits using the same old branch.

So your best 2 choices were either to hit that merge button before pushing the new edits to git. Or move your newest changes into a branch with a different name.

Anyway @atsju either way you want to deal with this is fine. We could just continue this same old PR with more changes in it. Let us know how you want to proceed.

[gr5]

Dale, regarding the compiler warning, there was a very similar warning in your code a week or two ago for the profile plot where you had a loop that went from -1 to 1 and this represented the left to right edge of the profile plot. You created a step in the for loop that was 2/(number of pixels in profile) and you multiplied the loop variable by the radius (in pixels).

I have seen a bug in profile plot that is caused by this double so we should probably take these compiler warnings somewhat seriously. I mean I think that is what causes the bug: If you have a wavefront that is an odd number of pixels (at least 601 works for the bug) then the profile when perfectly horizontal (at 0 degrees angle) will have one zero value at the end and the profile sharply dips up or down on the last pixel. It only happens at 0 rotation where the above mentioned for loop is getting a floating point off-by-one at the end of the loop. I'll create an "issue" now so we can refer to it. It's a pretty minor bug but would be nice to fix some day. By not using double in the for loop. Added: #333

[atsju]

I openned #334. It should be in same state as #333 yesterday. @gr5 you can probably just approve and merge it.

Best is probably to close #333 and open a new PR once #334 is merged (you can keep same branch if you want). Bu you can also just change title in #333.

[atsju]

Regarding the warnings with float I just want to share the rationale behind them with you. Then you can decide if it's needed to fix or not.
Floats are not exact numbers.
Example: Because 0.1f isn’t exact, x becomes 1.0000001f (or 0.99999994f) and the <= 1.0f check flips unpredictably across platforms/optimizations.
For loops can do more or less loops depending on rounding or optimizations.
== comparisons have to be avoided. They can fail because of small epsilon.

Fun fact 0 has several representations in float. But compiler takes care of that. This is why most of your ==0 are harmless when 0 is a manually entered value. It might not work if you do 0.4-0.3-0.1 == 0