Qc1 regression tests

Qc1 regression tests arising from the introduction of Qc2 flagging for useinfo.

The kvalobs database data from 200908 (taken from the backups of histkvalobs) was analysed and all the combinations of controlinfo and useinfo were extracted (approximately 2275 in total). The controlinfo values were processed by setUseFlags(const kvControlInfo& cinfo) with the existing Qc2 logic turned on. The following table displays the original controlinfo and useinfo as well as the new useinfo generated by the Qc2 aware setUseFlags. The fourth column indicates if there is a change to the useinfo or not. In approximately 37% of cases the useinfo generated was the same. useinfo(13) and useinfo(14) were overwritten by XX in the table since this is the userid set by HQC. AFter substituting XX it was found that the actual number of unique combination is 1076.

Questions:

• Is th a valid regression test?
• Are the differences a problem or not?
• i.e. are some differences only due to HQC activity and the Qc2 settings so far are ok?
• can we commit the updated setUseFlags back into kvDataFlag.cc?

The results of the regression test:useinfo-flag-test-200908.csv.gz

And here is an extract comprising 100 results picked at random from the total set of results (note the table is too big to load as one into the Wiki):

$ ~/DATA/FlagRegression$ awk 'BEGIN {srand()} {printf "%05.0f %s \n",rand()*99999, $0; }' wikitable.txt | sort -n | head -100 | sed 's/^[0-9]*//'

This is the same as above except the useinfo is initialised with the original value (before it was not set), still see the problem with useinfo(7)!!!!