This can be resolved by zooming in with the scroll wheel as far as possible.

I can confirm this fixes the clamp function in all bevel modes except for 'percent' - in that case the bevel still sometimes stops short.

The problem with inconcistancy when the selected face have an ended sharp edge is solved from what I can see :-D Thank you!

When I try to run Set From Faces on faces with a sharp edge I don't get the same result as I would if I run them separatly.

Ok, I will gladly test the other tools aswell :-)

Thank you! I've fixed both the problems. I would really appreciate if you could test out the other tools as well. Eg average with binary weights gives similar results to set normals from faces but also takes care of the cases of sharp edges.

I also made a rightclickselect proposal :
https://rightclickselect.com/p/modeling/Lvbbbc/improve-3d-cursor-snapping-consistency-and-control

Snap cursor to selected as well as Transform > Origin to Geometery both use the bounding box of the object rather than the barycentre of the vertices. This is not a bug but it would be cool to have the ability to use the barycentre instead.

The fix just committed only fixes this bug if the 'Clamp overlap' option is enabled. The fact that sometimes the existing code stops the edge slides when it hits another vertex was not something that was intended to find and fix all geometry collisions. Even 'Clamp overlap' doesn't do that, though it tries harder.

Thanks, so I guess we can consider the initial issue as fixed.

This is caused by a very small near-clip plane, in general if you have too small & large clip planes set it will cause issues with screen-space operations.

Confirmed, looks like some faces are not correctly split as separated items…

The inconsistancy problem seems to be when using the tool on faces that have a ending of a sharp edge.
I guess that the normal direction is based of the closest triangle? My expectation is that it's based of the entire face so that the normals of both corners match. Also it shouldn't matter IMHO if I run it several times on the same face but it does.

It is crashing when the mesh is edited so the loop normal spaces are wrongly invalidated. Nothing related to this tool, will fix by tomorrow. Could you provide an example when its giving inconsistent results? I've added sharp edges to the weighted normal modifier as well. I'll work on adding similar methodology to the tool if the modifier works fine. Also will make it so you can select both sides and perform the operation in one go.
Thank you,

It sometimes crashes when I have changed topology (triangulated ngons) and then doing a Set From Faces on the triangles but I have a really hard time getting a consistant case. I send you a screenshot of xcode trace so it might be I bit easier to narrow down?

True.
To be sure, I scaled all vertices on Z axis to zero before knife projecting.
Mesh analysis does show distortion.

I tested the following:
retested script on blender 2.79 from nightly buildbot.
BMESH fails -- get incorrect mesh ( mesh ends up with large, non-manifold hole ).
CARVE works, no crash on OSX.

Thanks for the report, but advanced things like autosmooth are ignored on purpose in Sculpt mode, to avoid too much overhead and slowdown… So no bug here. ;)

I'll check into why its crashing. I'll check into the set_normal_from_faces code.
Is it crashing when you are using some other tool that I've not written? If so, then I can understand why its crashing.

One thing I've noticed is that I can't select faces on the both sides of a sharp and do the operation in one go. I have got some crashes but I need to narrow the case before reporting them. Thank you!

Thanks for your feedback! I think I've considered all cases now, would appreciate if you could test it out and report again :)

This issue was fixed by rB723f2b2.

Not entirely unexpected but still a bit disapointing. Changes are part of any creative work. We can still do workaround with datatransfer but custom normal editing is already cumbersome.

There is indeed something weird going on with projection or so.

Thanks for the report, but no bug here, this is known limitation of custom normals currently - do not change topology!

Ahh! Totaly missed the option. It works like that but I still have problem if I use the Set From Face on the other side of the sharp edge (cap face in this case) instead.

Hey,
using set normals to face with multiple selected faces sets the normals to their face normal average. This has been an inherent part of the tool. If you want the figure on the right, you have to select all the flat non beveled faces and then use this function, along with checking "Keep sharp edges"

@Emil Assarsson (metabaronen)
I've added this functionality, albeit in my official soc branch for now.
You can test it out by building my branch: soc-2017-normal-tools.