On 07/12/2010 04:15 PM, Jeff Spaleta wrote: > On Mon, Jul 12, 2010 at 11:34 AM, Przemek Klosowski > <przemek.klosowski@xxxxxxxx> wrote: >> It doesn't look good when the application-provided examples crash. I see >> that it's a complex issue between the upstream and packagers, but I >> think the policy should be "it either runs or gives a clear message that >> it's unsupported" > > > What text do you need to read and where do you need it to exist in > order to make it _clear_ that the matplotlib examples provided as > documentation may require users to at a minimum to reconfigure their > matplotlibrc and to install additional python modules and are not > explicitly required for normal matplotlib use? And as a result of all > this none of the examples are "supported" (whatever that actually > means in this context) and not expected to work out of the box? > > I'll put whatever text is necessary to make it _clear_ that matplotlib > is a framework and not an end-user application and as such requires a > modicum of thought to use correctly. If you are unable to come up with > text that you feel meets the bar for clarity I can very easily just > not ship any of the examples and encourage users to use cut and paste > the online examples from the matplotlib website. .... > I'll go further and say that once I learn how to use AutoQA, and if it > allows it, I'll establish a subset of provided examples to use which > are expected to work out of the box for the default configuration and > run them as AutoQA tests for matplotlib. That's exactly what's needed, for matplotlib and other packages as well. Every file outside of this 'core examples' set could print the message "This example's functionality may depend on the specific configuration of matplotlib and other packages". Even better, could it be in some sort of try/catch block? -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel
