Tue. Nov 19th, 2019

Former Microsoft employees explain in detail why Windows 10 has so many problems

3 min read

Windows 10 system problems are generally relatively poor instability. Since the official release of Windows 10, the system has been experiencing frequent problems, and even the monthly security update has brought a lot of new problems.

Barnacules, who has worked for Microsoft for fifteen years, recently released a video commentary detailing the testing process of the build version of the Microsoft operating system team. It turns out that Microsoft has a very large test team. These test teams are divided into many different branches, and each branch has multiple different groups. This careful division ensures that all functional modules are fully tested, at least to ensure that each new feature has sufficient hands-on testing. These subtest teams meet every day to discuss the testing of new features and only merge the relevant features and code if there is no problem. It is worth noting that everyone on the test team needs to use the latest beta, which is installed on the physical machine as the daily main force of the test team.

Barnacules also mentioned that the original Microsoft test team also specifically set up a dedicated laboratory for testing CPU/GPU for Intel, AMD, and NVIDIA. These specialized labs are used to test for new builds or functional modules with compatibility or performance issues with critical hardware such as processors and graphics cards. The test team responsible for these laboratories will also interface with the manufacturer, so if the test team finds something, it can quickly determine and develop a solution. After the development team fixes the problem, it will be handed over to the test team for testing. After the test team passes the test, the code of the repair plan will be merged into the main thread.

In the video commentary, Barnacules also discussed the Windows 10 test project. Simply speaking, the test project does not help Microsoft solve too many problems. The main reason is that most beta users will not actively give feedback to Microsoft when they encounter problems. Of course, even if you give feedback to Microsoft, the final result may not be taken care of. The main reason for this is the dump log, which constantly logs various conditions and generates extremely large dump logs for analysis.

However, the actual situation is that the dump log will only record its details when the system crashes. Other “small problem” systems will not record the dump log. The full dump log is quite large, probably in the tens of gigabytes, hundreds of gigabytes, and possibly terabytes. Obviously most users don’t have as much space to store. That is, even if the user activity feeds back to Microsoft and provides a dump log, the actual provision is only a partial fragment rather than a complete running log of the entire operating system.

For Microsoft engineers, it is difficult to troubleshoot the problem itself through the fragment log. Over time, Windows 10 has accumulated more and more feedback. Then Microsoft engineers may have to give up selectively with so many problems. Windows 10 1809 lost file feedback is a typical example. When Microsoft engineers responded less and less to feedback from users, this, in turn, reduced the enthusiasm of beta users and gave up feedback to Microsoft.