Rescan command fails with weighted events

[Juergen Reuter]

Reported by FB.

[fbach]

Given a weighted event sample and a corresponding analysis with histograms (cf. foo.sin), if you rescan the sample and rerun the analysis, the original event weights are simply used again, thus producing identical histograms after rescanning.

[kilian]

I'll look into this. The alt values in the LHE output appear to reflect the change of PDF, so how should this be handled regarding analysis ...

[kilian]

OK, the sample Sindarin file uses alt_setup, and the behavior is correct. In the absence of array constructs within Sindarin expressions, it is not possible to plot the alternate weight. It is written to the event file as required, so no problem there.

However, I don't get a recalculation when the primary environment is changed. This is probably incorrect.

[kilian]

First of all, the sample input file contains several errors (or misconceptions) about the way the rescan command works. The alt_setup clause works correctly, but the calculated alternative weights appear only in the LHEF output, they can't be used for the internal analysis.

What does work is a straightforward rescan without alt_setup. At least, after clearing up the code and fixing various things, r6454. One has to set update_sqme and update_weight, and one has to make sure that the analysis definitions are valid during the rescan. I attach an updated example file.