bug 972970 went live last week, which means we can now scrape the set of symbols we actually used in processing out of the processed JSON. I'd like to see a report on this over a day's worth of crashes to see how many unique symbols we actually use, so we can get a feel for future symbol storage options. The query would be something like: select module.debug_file, module.debug_id from processed JSON's "modules" list where module.loaded_symbols == true It might be interesting to sum duplicates so that we get a count of accesses for each symbol file as well. The modules list in the processed JSON looks like this: "modules" : [ { "base_addr" : "0x8048000", "debug_file" : "firefox", "debug_id" : "5DAC464C5960263D3B5A909E275EEB2F0", "end_addr" : "0x8060000", "filename" : "firefox", "loaded_symbols" : true, "version" : "" },

Query used: WITH modules AS ( SELECT json_array_elements(processed_crash->'json_dump'->'modules') AS modules FROM processed_crashes WHERE date_processed BETWEEN '2014-03-05' AND '2014-03-06' ) SELECT (modules->'debug_file')::text AS debug_file , (modules->'debug_id')::text AS debug_id , count(*) FROM modules WHERE (modules->'loaded_symbols')::text = 'true' GROUP BY debug_file , debug_id ORDER BY count(*) DESC

Comment on attachment 8387869 [details] bug981021-result.txt This looks pretty plausible, since that top xul.pdb is from the current Firefox release: $ grep xul firefox-27.0.1-WINNT-20140212131424-x86-symbols.txt xul.pdb/EB6EB7EFD10D4611991C030F94F8ECED2/xul.sym Can you produce this in CSV form?

OK here it is as CSV - since it took a couple hours for postgres to spit this out, so I just converted the previous output to CSV. I made sure that python's CSV module can import it OK, so should work in excel etc. too

Thanks!