The case of the vector with an impossibly large size | Hacker News Reader

The case of the vector with an impossibly large size

devblogs.microsoft.com

74 pointssigna1136 comments

A favorite past-time of HN when a post with an interesting title on the microsoft.com domain appears, is to guess whether it's Raymond Chen, and then to fawn over Raymond Chen. All very well deserved, but are there any examples of interesting posts on microsoft.com from someone who isn't Raymond Chen discussed here?

It’s a pretty easy guess at this point since there isn’t much else of value published on that domain.

The MSDN documents on the website have been invaluable to me. I mean I started by writing chat tools on Windows and their API docs have been excellent in getting me started. I mean stuff like https://learn.microsoft.com/en-us/windows/win32/api/winuser/... I cannot understand why you think such a large website with so much documentation is of no value!

Value in the context of Hacker News means something that is amusing, insightful, or impressive. It is very rare for plain ol' documentation to make it to the front page.

Eric Lippert used to publish some good stuff, and Michael Kaplan.

Occasionally there is decent technical content from the SQL Server team [1].

[1] https://cloudblogs.microsoft.com/sqlserver/2022/07/21/improv...

Note that Michael Kaplan's original blog ceased to exist after he died and MS "renewed" MSDN blogs, but the archive is still available [1].

[1] https://archives.miloush.net/michkap/archive/

Eric is no longer with Microsoft, by the way. He was previously on the C# compiler team. I believe he was at JetBrains when I last looked him up. Looks like his blog "Fabulous Adventures in Coding" is on his website now:

https://ericlippert.com/

(I worked with him briefly in, I think it was, 2005.)

Was Scott Hanselman ever officially on microsoft.com?

Microsoft ships a moderately interesting modern language C# and so there are posts which attract a small amount of comment about new C# versions and their features, e.g 12 comments about C# 8.0

And likewise they ship new hardware sometimes, a new Surface Pro got hundreds of comments.

A big problem for these posts and for The Old New Thing (Raymond Chen's blog) is that Microsoft doesn't understand URL stability and so old material is too often moved, misfiled, or just destroyed altogether for no good reason. In particular it seems as though the web site (more vital now than ever for the company's appearance to the rest of the world) is routinely offered as a fun toy for a new executive to make their mark on to show how they've got exciting new visions for the company. Maybe this week Microsoft decided they loved blogs, everything is a blog now, want to announce Microsoft Boring Product 8.6.4? Write a blog post, and your customers can respond to the blog post, because we're an exciting modern tech compan... no, we abandoned that idea, all blogs are to be removed and the customer posts go with thme. Yes that means you've written what is now an article "responding" to customer questions which no longer exist but too bad we don't have time to fix that because now Microsoft is a video-first company, they are dynamic and hot, here are six hour long videos about Microsoft Boring Product and we are going to be delivering a new video every single week uh, wait, that's expensive, OK, no sorry there are no further videos about Boring Product, it's now a Wiki! Yes, experts even from outside Microsoft can contribute valuable material about Boring Product, we're going to transition all our boring stale material into user editable dyna... Nope! New plan, we're locking every wiki article and we're removing the unofficial content, yes, customers were relying on BigBabo4239's sixty paragraph explanation of how to use Power Shell with Boring Product but that no longer aligns with our corporate strategy so we've deleted it...

> the web site [..] is routinely offered as a fun toy for a new executive to make their mark on to show how they've got exciting new visions for the company

To be fair to Microsoft, this issue affects a great many other companies, even outside the technology space.

My personal favourite has got to be www.britishairways.com, where you can discover older and older versions of their website layered like the concentric rings of some ancient tree.

https://www.britishairways.com/travel/home/public/en_gb/

https://www.britishairways.com/travel/webforms/public/en_gb?...

and even the gloriously broken

https://www.britishairways.com/onbusiness/how.shtml

While the British Airways link is amusing, it's much less amusing when you've been linked to a now moved documentation page, or worse, a now deleted download page.

> older and older versions of their website layered like the concentric rings

I like this approach. Keep old content forever, just mark it as "old" at the top of each page and archive it as a static html page at its original URL.

(archiving as static html needs to be done at some point, otherwise you'll still be running php4...)

See also Windows setting menus. People see this as a flaw, but I much prefer that to breaking existing functionality for the sake of a redesign. Make a new interface for simple usecases, but keep the old one around just in case.

I feel like this is the norm at big companies. Oh, noone is using platform abc any more, it is too basic. We must switch to def! And we can't afford the retain old links feature, who needs that anyway.

I've seen platforms where if no content author was around to vouch for old technical documents, they'd get summarily deleted. Saves storage! sigh

Stephen Toub

https://devblogs.microsoft.com/dotnet/how-async-await-really...

Yes, papers or articles from MS research, e.g. from Simon Peyton-Jones or about Koka or Lean. https://news.ycombinator.com/item?id=36471591 ... Guess something about VS or VS code too.

Busted. Saw this and immediately thought it had to be a Raymond Chen post. Whatever he’s getting paid, it probably isn’t enough.

I've been out of C++ for a couple of years now, but std::call_once is totally new to me, despite being in C++11.

It really is a language where if you got 5 different people to write a solution to something you'd get 5 wildly different answers.

Six, at least. Also at least one of the answers the five people swear they use will turn out to be either Undefined Behaviour or a quirk of one particular compiler/ toolchain.

Like sure, volatile integers are basically just atomic variables with easier syntax... if you are on x86-64 and using MSVC. Otherwise no they really aren't.

A nice, related meme:

https://mikelui.io/img/c++_init_forest.gif

And on that subject https://leanpub.com/cppinitbook - a book just about initialization in C++ which is in some sense (more pages) longer than my edition of K&R ("The C Programming Language").

Despite all this complexity, C++ 23 has problems with mistakenly uninitialized variables and still lacks a coherent strategy to do deferred initialization correctly.

That GIF should mention the relevant C++ standard, because next year there will be at least one more additional way to initialize.

I consider that a feature tho.

After working on golang projects for a while, I really like having one obvious way to do things.

Is it habitual for Microsoft to help their customers like this? Is there some kind of support or consultancy contract for it?

Microsoft has probably the largest support organization of any software company. A lot of it is outsourced support that mostly runs through support trees, but enterprises can and do pay for support from actual employees who can actually debug issues and involve developers when needed.

It’s never lupus. It’s always race conditions.

It's interesting that people have such trivial problems with their C++ code and still manage to get an expert from Microsoft to look at it for them.

Also what I find interesting is that modern tools like sanitizers would quickly say what's going on without needing any human analysis.

It's Raymond Chen.

It's not that the customer deserves Raymond Chen. Not even HN deserves Raymond Chen.

It's just that Raymond Chen is once again gracing us with his vast and deep insights.

Legend.

The fix is kind of weird, no? If the lazy initialization of the widgets fail, won't it stay stuck as failed forever (i.e. until the containing class is destroyed)? Is this intended behavior?

No, at least not if exceptions are propagated properly. The flag that determines if call_once should actually run or not doesn't get flipped if there's an exception.

I missed it. How big was the vector

> The m_widgets is expected to have only four widgets, but it somehow found itself with ten, and sometimes as many as a hundred widgets.

Asking that question is UB.

What does UB mean

Reply on news.ycombinator.com