Number: 1949
Title: Bug using sub graphs
Submitter: carlos Libório
Date: Tue May 11 07:15:41 2010
Subsys: Dot
Version: 2.26.3
System: x86-Windows-XP
Severity: critical
Problem:
Dear Sirs.

I've been having a quite unfamiliar problem using graphviz. I'm building an application that uses graphviz to design some process models.

When i started to use subgraphs, sometimes graphviz crashes without any "possible debug" message.

I've trying to figure out what the problem is, but what i found out is that if i just delete some of the lines of my deployed dot file, without any known logic, it works, so i can't find any logic concerning this error. I've tried to update to your last version but the problem persists.

Is it possible to try a help me on this ?

Best regards,

Carlos Libório
Input:

<private>
Comments:
Please keep my input private.

[ellson] Carlos asks for his graph to be kept private. Could you remove it from the bug system, please?

Carlos,

I worked your graph a little, so that at least it generates something:


    - I moved all edge declarations outside of the clusters.
    - I removed a number of duplicate node declarations.
    - I moved a few nodes into, or out of, clusters where it seemed to make sense, but I'm sure I got it wrong.
    - I added null labels to Initialization and End, since the fixed size leaves no room for the text.

There doesn't seem to be a good flow to your graph for a nice looking directed layout with dot. Hopefully this is something you can fix now that you can see the output.

I tried using the "fdp" layout engine, which also supports clusters, but the result doesn't look any better.

Hope this helps?

[ellson]


> I take it the graph does not crash on Linux?
>
> Carlos,
>
>   What output do you get if you run dot -v?
>
>     Emden

No, the original crashes on Linux. It crashes with:


    Warning: node 'Initialization', graph 'G' size too small for label
    Warning: node 'End', graph 'G' size too small for label
    Error: trouble in init_rank
        virtual 1
        virtual 3
        ...
    Aborted (core dumped)

[erg] Then this is another case of the "trouble in init_rank" bug. The only question then is why he didn't get the error message when he ran dot.
Owner: *
Status: *