On Jan 17, 2014, at 9:56 AM, Beman Dawes wrote:
We have had a long term problem with regression test reporting going offline unexpectedly. The process used XSLT processing was hard to configure, horribly slow, and very fragile, and those factors made it a pain to run.
Last year, Steven Watanabe rewrote the report generation in C++, and those issues have pretty much disappeared, so the process is now far easier to run. It runs in five minutes or less.
Hooray C++! Hooray Steven! And Hooray Tom Kent who runs the reports!
The remaining reliability factors are local ones like system crashes. Since the reports run as cron jobs, it can be quite a while before anyone is inconvenienced enough to report it to the Boost list, and Murphy's Law ensures that it is always the day after the person running the reports left for a week-long back country hike.
We can mitigate that by having several people running the reports. I've written some docs, so it shouldn't be too hard, although right now it does require a *nix-like system. I've been running them on a Windows host using a VirtualBox virtual Ubuntu system, and have found it to be easy.
One possibility would be for Tom to continue to run the reports on the hour, a second person to run them twenty minutes after the hour, and a third to run forty minutes after the hour.
Please volunteer to help make regression reporting more timely and reliable!
I can help with the reporting if you still need some volunteers. This would be on RHEL in case that matters. -- Noel