Problem 1177

Summary:	Visualization of G4Polyhedra with HepRepFile and OGLIX gets stuck in g4.9.4
Product:	Geant4	Reporter:	Alex <alexjaguar>
Component:	visualization	Assignee:	perl
Status:	RESOLVED LATER
Severity:	normal
Priority:	P5
Version:	9.4
Hardware:	PC
OS:	Linux
Attachments:	The union of 5 rails A single rail zoomed Source code of a simple test program to reproduce the bug

Description Alex 2011-02-23 11:28:52 CET

Hello.

I have some G4Polyhedra in my geometry, also I use a box subtraction from the polyhedron. I used to visualize my geometry using HepRepFile system. It worked fine in g4.9.3 (up to patch-02) and now in 4.9.4 it gets stuck with no messages. I tried to trace the problem with gdb/ddd.

I end up somewhere around HepPolyhedronProcessor::execute1. Then I tried also OGLIX and this time it was somewhere around BooleanProcessor::takePolyhedron(). I didn't try the other systems. These functions are called many times (constantly?). I think it comes somehow from the recursive boolean processing announced in 4.9.4. I have to mention that the event simulation seems to work fine, no warnings/stuck.

This is all the information I have at the moment. Unfortunately, I don't have time to dig even deeper. I would appreciate if you can give a hint where the problems can come from, or if you are not able to reproduce the problem, then I could try to provide a small test program that gets stuck (not sure how much time would take this though).

Thank you very much.

Comment 1 Alex 2011-04-11 17:17:38 CEST

Any news on this bug? It still remains in G4.9.4.p01.

Comment 2 Alex 2011-11-08 11:03:27 CET

Created attachment 139 [details]
The union of 5 rails

Comment 3 Alex 2011-11-08 11:04:17 CET

Created attachment 140 [details]
A single rail zoomed

Comment 4 Alex 2011-11-08 11:05:45 CET

Created attachment 141 [details]
Source code of a simple test program to reproduce the bug

Comment 5 Alex 2011-11-08 11:07:01 CET

Hello again,

I tried to dig deeper and found out the source of the problem. We have series of solid unions. The first union is of the rail shape (three united boxes). The 4 of rails are positions in parallel, and then 8 more perpendicular with intersections. It creates a complicated volume and as far as I understood this solid is internally converted to G4Polyhedra for visualization. I guess there appear too many vertices/edges that the BooleanProcessor goes to a very long loop, and cannot handle it (or at least does it very slow). However, if I put 4 parallel and 1 perpendicular rail, it works. Putting 2 perpendicular may work, I did not wait enough, but this outputs warnings:

BooleanProcessor::createPolyhedron : too many edges

Maybe this is not a very good way to create solids, I simplified that in our geometry, but if you consider this as a problem in G4, then I created a very simple test program based on exampleN01+exampleN02. I made comments in src/ExN01DetectorConstruction.cc to explain what cases work and what don't. There is also a vis.mac macro to create a HepRep file. I also attach two pictures of
a) the single rail and
b) the union of 4+1 of them (the maximum that works).

Alex.

Comment 6 perl 2012-09-07 09:01:24 CEST

Because we are aware that the boolean visualization is imperfect, we will save this as a test case for whatever future time when we have sufficient resources to significantly revise the boolean visualization. We thank the user for providing a clear test case.