Hacker News Reader

Comment by copper_think | Hacker News Reader

One thing not mentioned is that Visual C++ and Visual Basic historically were separate IDEs with separate codebases. When the time came to unify them, only one of them could continue on. My understanding is that Visual Basic won, and that today's Visual Studio IDE (devenv.exe, msenv.dll, etc.) is the continuation of that VB codebase.

I don't actually know in which release that transition happened. But since there's a screenshot of each version in the article, presumably that transition is visually documented...

View on HN

That might have been the transition from Visual C++/Basic 4 to 6 (I skipped 5), but the cataclysmic one was the switch from Visual Studio 6 to Visual Studio .NET, when large portions of the IDE and build system were rewritten in .NET. Visual Studio .NET (2002) was much slower and much buggier than VS6. The native debugger was glacial at conditional breakpoints and debug output, the build system took an eternity to do a dependency check, its UI visibly redrew more slowly, etc. It was so bad that Microsoft had to create a special offer for the Visual Studio .NET (2003) upgrade for only ~$30.

This transition was not great for Visual Basic developers either since their language was transitioned from generating native code (VB6) to becoming dependent upon the .NET Framework (VB.NET), supported secondarily to C#.

To this day I would consider VB 6 the epitome of COM development tooling, neither .NET, nor the UWP trainwreck were as good, and C++ tooling only MFC is worth of mention, as something putting development experience first, instead of Windows division politics.

Quite awkward for what is such a key technology on Windows.

If memory serves, it was Visual InterDev that won, actually. The VID IDE was the forerunner to Visual Studio.NET 2002, which was the first unified Visual Studio IDE.

n.b. I worked on the Visual Studio Core team, which maintained devenv.exe, among other things, from 2003-2007'ish.

Visual InterDev was the first IDE I ever used outside of notepad. It was bundled in with one Microsofts early webediting tools with office (97/200)....but not FrontPage. I thought I was touching the future, it was so amazing! I really don't remember much about other than the impression that it left on me.

After Visual C++ 6. They broke a lot of the C++ IDE features and they weren’t as good as the prior versions (dialog editor etc) so for a long time we preferred staying on 6. I think if we could have the newer compilers but the snappiness of that UI many developers would be happy.

It’s also a product of the segmentation of the developer tools in Microsoft. The Windows team was responsible for the compiler rather than the Developer Tools team.

No, that's not really true. I was on the C++ compiler team from 1991 to 2006. When I first started, the DevTools team reported up through the Windows team, but never really felt a well-integrated part. We were never in the same building as the Windows team, for instance. I remember, probably 1992 or 1993, driving from building 4 where the compiler team lived to the Windows building (forget which one that was, maybe in the 12 to 15 block back then?) to get a copy of the Windows NT source on a hard drive. That's because I was a dev on the C++ compiler back-end team then (moved to the front-end in '95, IIRC), and compiling that source was a major test of the 32-bit compiler I was working on.

Don't remember when DevTools was re-orged out from under Windows, but I'm pretty sure it was by '95, and well before VC++ 6.

Did you ever work on the cross edition that would compile Windows apps for the Mac? I think that was a version 4 fork that never got another version.

No, there were two devs working on the 68k Mac compiler, with ~10 devs on the x86 side (though both targets shared a lot of code and differed mainly in the late codegen and peephole optimization phases). I never worked on the 16-bit code; the 32-bit and later 64-bit x86 backend was a different codebase from the 16-bit stuff.

If I remember from poking around at the 16 bit version in that timeframe (1.52?) there was still an option to generate P-code and (this is now nearly thirty years ago) that was used by Excel to minimize code size for the segmented 16 bit systems.

Any idea how many devs are working at Microsoft on the C and C++ compilers these days? I've heard rumours that there's more on the Rust team and that C++ is taking a back seat.

No clue. I left as a full-time employee in 2007, did a few contractor gigs with various old teams of mine to help out, but that was done by 2013. I lost touch with how things were going internally after that.

The dialog editor is a good example of the damage that was done with the VS.NET transition. When they rewrote it to use the WinForms-based UI, they introduced a fundamental bug: the Z-order for picking was reversed so that clicking on a stack of controls selected the one on the bottom. Very annoying for controls like group controls intended to be stacked below other controls. Bug filed, WONTFIX'd, and it's still broken to this day.

Afaik Win32 API explicitly states that overlapping sibling windows/controls is not supported.

No, overlapping sibling controls is absolutely supported in Win32. It has a defined Z-order for child controls along with functions to manage it:

> The system maintains the z-order in a single list. It adds windows to the z-order based on whether they are topmost windows, top-level windows, or child windows.

https://learn.microsoft.com/en-us/windows/win32/winmsg/windo...

The WS_CLIPSIBLINGS window style is also explicitly for controlling cross-clipping of child windows during painting:

> Clips child windows relative to each other; that is, when a particular child window receives a WM_PAINT message, the WS_CLIPSIBLINGS style clips all other overlapping child windows out of the region of the child window to be updated.

https://learn.microsoft.com/en-us/windows/win32/winmsg/windo...

Furthermore, group boxes and tab controls require overlapping child windows, because they are designed for the controls that visually appear within them to be sibling controls that are raised in Z-order. This is why the dialog editor's current behavior is broken, because the editor's picking behavior is opposite the drawing behavior and USER32's hit test order.

Right! It is wxWidgets that doesn't support overlapping siblings due to other back-ends (besides Win32) lacking support.

the big thing I missed was their macro editor, it was really good for automating things.

I think that is right and perhaps that is why, as I recall, Visual C went from almost unusable to a pretty good development environment between version 2 and 4. This is all (old) memory and anecdote.

Reply on news.ycombinator.com