[Halld-offline] fix to make JEventSource_EVIOpp crashes on PS skims

[Richard Jones]

Hello Alex and all,

I just submitted a pull request that fixes the issue you had with crashes
in JEventSource_EVIOpp when reading from PS skim evio files. I have tested
it on both rawdata and PS skim files and it now behaves in a similar
fashion to the sequential JEventSource_EVIO. The problem is that PS skim
evio files are "seeded" here and there with sequences of zero words
interspersed between the events. Through no deliberate design feature that
I could discern, the old JEventSource_EVIO code (actually HDEVIO::read more
specifically) would survive by skipping over these, generating silent
warnings (unless EVIO:VERBOSE>0) and eventually find the next non-zero
word, which turns out in all cases that I found to be the beginning of the
next event.

I duplicated this behavior in the new code I added to HDEVIO::ReadNonBlock
(actually HDEVIO::MapEvents called from ReadNonBlock) but did not silence
the warnings because they may indicate loss of information somewhere in the
skim generator and I don't think we can ignore it forever. At least now it
doesn't crash.

I also fixed the infinite loop in JEventSource_EVIOpp. I called the pull
request pp_evio_fixes_rtj or something like that.

-Richard Jones
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-offline/attachments/20160906/cf32e560/attachment.html>