Number: 819
Title: CVS build of neato dies with "caught SIGFPE e" on even the simplest graph
Submitter: Frank Gibbons
Date: Tue Nov 15 13:27:21 2005
Subsys: Neato
Version: 2.2.1
System: x86-Linux-2.4.18-686-smp
Severity: critical
Just try running neato on the input file below. It doesn't appear to matter what specifically the input is, it always dies.

I'm upgrading from 1.14 to 2.2, building from the CVS sources. I'm really interested in the TopFish visualization technique, but also in just laying out small graphs (up to 100 nodes), to understand the protein interaction network in humans. GraphViz is still the best tool I've come across for quickly laying out that kind of graph, and allowing me to customization the decoration according to various kinds of biological information. I've been using graphviz for a number of years (Stephen North even wrote us a letter of support for a grant application a year or two back - that was in connection with using graphviz to visualize multigraphs). I'd be really glad to have some kind of fix.

I noticed from reading the earlier bug reports, that I'm not the only one who encountered this problem. Like the previous user (bug #595), I'm using gcc 2.95.x (x=4 in my case), I'm not really interested in upgrading, so looking for a fix.

Thanks in advance,

-Frank Gibbons

graph g {
two -- one
[north] Frank: topfish (er, topview) is not in graphviz, but we would like to find academic partners who are interested in working with it. We can issue university ("non-commercial) licenses but it's not open source.

With gcc problems, you might try compiling everything without -O and if that does the trick, then figure out which file in particular tickles the gcc bug. I can't remember the likely suspects any more (at least for neato; in dot it was generally the spline fitting functions). It's been a long time since I followed the area but there used to be alternatives to gcc, such as lcc and of course if you want to spend real money there's Comeau, Intel, etc.
Owner: *
Status: *