WinDrvr - All Comments

re: Improving the documentation

Jennifer Stepler — Fri, 25 Jul 2008 16:17:50 GMT

Just a note that the WDK docs are also available on the MSDN website. Each page of the MSDN docs allow the reader to rate it, post comments, or post additional information via a "Wiki". The documenation team is very proactive about using the ratings, comments and Wiki entries to improve the docs.

re: Fixing WinHEC

Thomas F. Divine — Sat, 19 Jan 2008 23:43:04 GMT

Personally, events specifically for driver developers are ideal. The DevCon format, held on the Microsoft campus makes the most sense for driver developers.

re: Tag, you're it

Volodymyr Shcherbyna — Tue, 11 Dec 2007 13:56:45 GMT

Don,

If I recall correctly, the verifier.exe will halt a bug check in case if you specify a wrong tag value for ExFreePoolWithTag. So, at any case, if you pass driver verifier test you can be absolutely relaxed that you do correct things.

re: Coding Guidelines

Volodymyr Shcherbyna — Tue, 11 Dec 2007 13:53:01 GMT

I wonder, what was the name of that toolkit.

re: Coding Guidelines

Ayush Gupta — Mon, 23 Jul 2007 20:09:34 GMT

Coding style:

It would be great if you could provide some pointers to it.

Should we follow some standards specific to the driver code?

Are there any "Best practices" for the same?

you can contact me at:

ayush.gupta@yahoo.com

re: Bleeding edge and far from it

Steve Dispensa — Sat, 02 Jun 2007 17:42:55 GMT

I think DTM is based on the internal Microsoft WTT (windows test tools?), minus some stuff like integration with their bug database and such.

If memory serves from the MVP summit, there is a group on the testing side that put the testing framework together, along with the device simulation framework. Both look interesting, but I certainly don't have time to mess with COM.

re: How tuned is your time machine?

Steve Dispensa — Thu, 31 May 2007 22:13:55 GMT

Hear! hear! I wish people took source control more seriously. It has saved me often.

SourceSafe is too painful for words, though; I have had good luck with:

- CVS, for many years, although it's old and showing its age

- Subversion, which is a better CVS, mostly

- Perforce, which I use at work, and which is fantastic.

The first two are open-source, and Perforce is free for 2 users, so if you're a small dev shop, it might work for you. It gets expensive after that, though - $800/user or so.

Commit early and often. Branch for releases. Tag for patches. Your life will be happier.

re: Crossing the Undocumented Line

Harry Tasker — Tue, 29 May 2007 16:27:04 GMT

"As a consultant who has more than once taken on projects Microsoft has said are impossible, "

This is interesting - a few questions though:

Did Microsoft say the entire project was impossible or part of the technique it's based on? e.g. Did they say you just cannot have a product like this work reliably on Windows OR You cannot make this product work reliably using this-technique-you-mention because it's undocumented (and the using the documented way won't give you the required benefit - e.g. performance - etc.?

In other words, was it because Microsoft didn't want to bless the usage of undocumented calls or was it because Microsoft guys didn't think of the technique you used as a viable approach for that project?

Can you also blog about how you perform the test in such cases? Do you consider all the combinations with various service packs and updates? I've seen that some updates combinations can break something especially if the administrators don't issue 'recommended' updates, but only install 'security' / 'vulnerability' updates.

re: Bleeding edge and far from it

Thomas F. Divine — Mon, 21 May 2007 15:11:22 GMT

Sorry I missed WinHEC and a chance to see you. :-( I agree with you that some recent MS tools for driver developers seem from another world altogether. Apparently the Windows Driver Testing Framework falls into the category. I believe that the Windows Driver Testing Framework is the foundation of the DTM, and I can certainly say from several months of misery and frustration with DTM that it is not a solid foundation. If a test failed with the legacy HCT a developer knew the it was likely that his or her driver was faulty. If a test fails with the DTM it is _possible_ that the driver is faulty, but also likely that the DTM framework or the DTM test being performed has a bug. The "KISS" principle ("Keep it simple, stupid!") is a key cornerstone of driver development and testing. Some folks at Redmond are loosing sight of this.