gnuplot locks up plotting log(x**0.5)/log(x)

It's hard to imagine what a high importance bug for a plotting application might be if being unable to plot a simple, smooth function results in a lock up is considered low.

According to https://wiki.ubuntu.com/Bugs/Importance, Low bugs are those which affect functionality, but to a lesser extent than most bugs. While you may consider this a simple, smooth function, it requires specifically using 'log(x**0.5)/log(x)'. Changing this to log(x**0.6)/log(x) works perfectly fine. A 'high' bug would have to have a severe impact on the application, and this appears very moderate to me.

I've also reported this upstream.

yes, it is the sqrt operation which causes the problem. To me it suggests a more serious bug in the parser/compiler.

If it appears very moderate to you, why was it marked low? Surely that would be 'Medium'?

Thank you for upstreaming the bug.

Please see https://wiki.ubuntu.com/Bugs/Importance. I categorized this bug as 'A bug that has a moderate impact on a non-core application'. You could not say that this is 'A bug that has a severe impact on a non-core application. ', because that would mean it is largely unusable, which is definitely is not.

Would it be possible to link this bug with the upstream one?

gnuplot hasn't been registered as a project on Launchpad yet, which makes that not possible.

gnuplot has a bug tracker at http://sourceforge.net/projects/gnuplot/ (bugs can be linked from there to Launchpad).

That's correct, but there is no project on Launchpad for gnuplot, which makes it not possible to create a bug watch. It is also specified to report bugs to the <email address hidden> mailing list in the gnuplot documentation, though I'm sure using the sourceforge bug tracker is fine.

I can't reproduce this in gnuplot_4.4.0-1. The function plots just fine, albeit with a small amount of noise from machine precision limits.

@njh - Does this still affect you with the new gnuplot package?

Yes, I concur, this bug is fixed! Thanks everyone for your efforts.

I spoke too soon. I can in fact reproduce the same hang with log(x**k)/log(x) for k in [2, 10, 100], perhaps minor changes have made the trigger case move around a little.

Now this is weird... I just tried this again, using an i386 machine (I tested on amd64 yesterday) and I can confirm the original bug.

@njh - which platform are you using, please?

Forwarded to https://sourceforge.net/tracker/?func=detail&aid=3085168&group_id=2055&atid=102055

I'm using an AMD Athlon(tm) 64 Processor 3800+ ubuntu 10.04

OK, which kernel architecture do you have? If you're not sure, please type "uname -a" in a terminal.