• The saga of Microsoft testing

    Buried in a Reddit thread, there’s an interesting (and, I believe, accurate) retelling of the recent history of Microsoft testing. Here are the high points:

    3MuchCaffeine5Me:

    Microsoft converted all of the Software Development Engineers in Test (the software engineers who write automated test software) to full SDEs (Software Development Engineers).

    I have full confidence that the SDETs are doing well as SDEs. After all, it’s an SDE, just writing domain specific software. Same skills and what not. The former SDETs are probably the highest quality engineers on Microsoft’s team. That specialization enforces a REAL drive in you to create quality software.

    What I don’t have confidence in is that any developer at Microsoft is able to fully test their own software to meet deadlines without having to fudge tests and just push things through. Nor able to create maintainable automated tests on top of writing regular consumer software. Likely, deadlines are forcing them to use recording tools to make automated tests. Tool generated automated tests are extremely flaky and no where near as good as writing one by hand, as a dedicated SDET would do.

    As you can see, we ended up with the buggy cluster that is the current situation in Windows 10.

    Forest-G-Nome:

    Microsoft laid off their QA in 2012 and outsourced a very minor compliance group to VOLT.

    I know, because I was part of that group.

    When it comes to SDE/SDET at microsoft, they basically make sure builds can be built, not that they meet quality assurance standards.

    3MuchCaffeine5Me:

    They have less resources to devote to testing. The problem would be that these things could be getting thrown past deadlines without receiving as much vetting as they would have had in years past. Software is buggy. Designs are imperfect. Bugs are not a chance happening but an expected occurrence. If you don’t have the proper resources to find them, your customers will suffer.

    You can have all the developers in the world. If you have barely anyone testing it, or barely anyone caring about proper testing because delaying a deadline would make them look bad for not delivering on time, then you’re going to have a rotten piece of Swiss cheese instead of working software.

    RequiemDK:

    Their idea of “having people testing their software” is probably the insiders programme, which looks good on paper but really effin’ dubious when it comes to quality and workflow.

    It’s not only about software too. Their Surface QA is the pits. Just look at the number of hardware issues that the SP4 had and how long it took MS to even acknowledge them, and the issues that current Surfaces continue to have with the pen.

    Added, without snark: I think they’re right. – Woody