September proved to be an interesting month in the DevPartner lab. After the descent phase into the GA distributions of the DevPartner Studio 10.5.2 English and DevPartner Studio 10.5.2 Japanese releases, we tackled one feedback path for bug sources that we had wanted to reestablish for months, actually years. That feedback path is called Windows Error Reporting, or WinQual, a very handy service for applications on Microsoft platforms to report crash and hang data. I say this was on my TODO list for a very long time because until now, DPS 8.0 was the last release with module mapping pushed to WinQual, and the log jam on paperwork that I started in March finally reached a breakthrough in September. We were able to map our modules and establish bug reporting stats for the three fielded DPS releases in the 10.5.x series. The data provides some interesting insight into where crashes are happening around the globe. In complete candor, as hard as we may try to make our product bulletproof, because its very nature is to ride along underneath potentially buggy, partially finished code under test, inevitably we will crash for someone somewhere (but I pray please not for everyone everywhere!).
The data shows a very nice trend over the last 90 days that overall crash and hang events reduced from around 1000 total for DPS 10.5.0, fewer than 300 in DPS 10.5.1, and fewer than 100 in DPS 10.5.2. The Japanese release fares best with only 11 instances in the whole Japanese region since release. I’m encouraged by the positive trend. I’m also a realist: DPS 10.5.0 has been in the market for nine months, 10.5.1 for six months, and 10.5.2 for just two months, so crash rate is falling, but uptake of 10.5.2 has really only just begun. However, the data also shows the concentration of crashes is not so pronounced in the later dot releases. In the 10.5.0 release, BoundsChecker modules show up most often, whereas in the later releases instances are occurring fairly evenly all around the suite. And I convey most sincere apologies to whoever must have run DPS 10.5.0 compile time instrumentation in a death loop around August 26; specifically NMCL2.EXE, our backend instrumentation engine, crashed 349 times in the span of 48 hours. Fortunately that bug is either fixed in the later dot releases (we hope!), or someone killed their script for good and never resumed it (sorry!).
The one good aspect I must convey regarding our approach to stability is that we have been diligent about refreshing the GA build as often as we possibly can. We’re continually sweeping feedback from customers into the product as fixes, stabilizations, and performance optimizations with every major, minor, and dot release. In our next release, we will be putting in a product update notification feature, such that even if our email bombardments, In Focus newsletters, and SupportLine notes don’t reach you, DPS itself will give you an occasional tap on the shoulder that a new build is ready for you. And hopefully with ever fewer crash instances inside, too.