Windows 10 update traps some systems in a boot loop, Microsoft promises fix [Updated]

[Old Account]

Microsoft insists that this problem only affected people in the Windows Insider Program, though it has not explained why others not in that program would have been impacted by the latest patch. A fix has been pushed out for the issues but the company has provided no details into what went wrong or what the patch fixes. The supposed link to a knowledge base (KB) article that’s supposed to describe the problem is actually dead as of this writing.

Original story below:

For well over a year, Microsoft has tried to position Windows 10 as a new type of operating system — one that’s continually updated without allowing for meaningful customer intervention, with security, driver, and software updates combined together by default and updated on Microsoft’s timeline. Pro customers can delay updates for a period of time, but non-enterprise users aren’t allowed to push them off indefinitely.

The company has repeatedly promised that by giving up this control, its customers will receive more timely updates, features, and improvements. It’s the kind of promise that requires an ironclad commitment to shipping stable software — and Microsoft has been dropping the ball in myriad ways ever since Windows 10 launched. Last week’s update for Windows 10 Anniversary Edition (KB3194496) has trapped some users in an unending reboot loop.

Microsoft has told ZDNet that it’s already in the process of fixing the problem and will release an updated script in the near future. The problem doesn’t affect all users, but issues like this often leave me wondering how machines trapped in an endless boot loop are supposed to get the fix in the first place. Presumably the company will release instructions on how to short-circuit the boot loop manually and apply the appropriate patch.

Part of why this is making waves is because the problem was reported by early adopters before the patch ever went live to the larger community. Microsoft apparently either ignored that feedback or never read it in the first place. Prior to 2014, Windows releases were evaluated by an extensive programmatic testing team within Microsoft itself, but Satya Nadella gutted this division when he became CEO. In its place, Microsoft rolled out a system in which developers were expected to troubleshoot their own code, with remaining QA resources dedicated to real-world tests and troubleshooting.

In theory, the above should’ve allowed Microsoft to roll updates more quickly and improve its software on a more rapid cadence. We have seen new Windows 10 features appear more rapidly than previous iterations of the operating system. Along with those benefits, however, has come a series of problems. For example, Windows 10 Anniversary Edition broke most webcams because it no longer supports compressed data streams, which the vast majority of webcams rely on in order to function. At least some Kindle devices crashed Windows 10 AU when plugged in, and Microsoft’s KB3176934 brokePowerShell. Last year, the company pushed an update that removed software from users’ systems after erroneously flagging it as incompatible with Windows 10.

None of these issues are show-stoppers in and of themselves. There have always been Windows patches that caused problems on a small number of machines, and that’s inevitably going to happen when you have 400 million devices running any operating system given that Windows 10 can run on hardware that’s 10+ years old. But as John Dvorak points out on PCMag.com, this problem takes on an entirely different dimension given how Microsoft now forces people to take updates whether they like it or not. Youcan still prevent Windows 10 from updating, to be clear — the Windows Update service can be manually deactivated. But this is a brute-force solution to the problem that exposes users to significant security risks.

One potential explanation for how these problems keep slipping through is that Microsoftdoesn’t recommend Windows Insiders test fast ring deployments on primary systems or mission-critical hardware. This means that a lot of testing gets done from within a virtual machine. VMs are great for testing a number of scenarios and applications, but they can’t completely replace dedicated testing on production-ready hardware. VMs aren’t the sort of thing people tend to use for Skype or other peripheral tests.

Since Windows Insiders reported the latest problem before the patch went live, it suggests Microsoft needs to be paying more attention to the feedback it gets from beta testers, period. The company needs to lock down these issues before something truly catastrophic slips through the vetting process, or before the steady drip-drip-drip of flaws and failures builds up to the point that it threatens consumer perception of Windows as a stable operating system.

And yes — before anyone says anything snarky in the comments, I’d argue Windows isgenerally perceived as a stable operating system now. In the late 1990s, it wasn’t uncommon for people to reboot once or more per day. I considered it an accomplishment if I could keep my own Windows 98 SE box online for a week without needing to reboot to fix an OS instance that had slowed to a crawl or otherwise become unusable. Today, PC uptime is routinely measured in months, provided you don’t need to reboot to apply an update (or provided the machine doesn’t reboot itself after an automatic update). Other staples of late-90s and early 2000s computer advice, like the need to periodically reinstall your operating system to keep the PC running at an acceptable speed, have similarly vanished.

The way it stands now, Windows 10 risks shifting consumer perception in precisely the wrong direction. Ideally, Microsoft would decouple its security updates from its feature updates, keeping the former mandatory and letting the latter be optional. Failing that, the company needs to invest in solving these problems with a more robust QA strategy.

Now read: Windows 10: The best hidden features, tips, and tricks