The CAPITAL LETTERS trick that helped merge Windows 95 into NT

The early versions of Windows NT were the last hurrah for the Windows 3.1-esque Program Manager. But getting the Windows 95 shell into the codebase occasionally required using CAPITAL LETTERS. The development of Windows 95 was underway around the same time as Windows NT 3.1 was being prepared for launch, looking like Windows 3 …

[ChrisC]

Even with source control systems, being able to simply search the source code within your IDE/editor and almost instantly find every instance of an original vs a fixed version of a macro, function call etc, can be invaluable.

[Yet Another Anonymous coward]

Although it did lead to the years of having to start all your code with:

#define WINDOWS_LEAN_AND_MEAN

#define FOR_GODS_SAKE_TURN_OFF_THESE_STUPID_DEFINES

if you didn't want all the occurrences of "max" and "min" in your code to be silently turned into Windows macros

[JB1]

It's easy to scoff but the industry was a lot younger then and the pressure to deliver these products was sky high.

[wolfetone]

And to deliver them working, complete, on day of release.

Not now where you deliver something which essentially downloads a 50GB patch on day one to make the damn thing work.

[Bill Gray]

I had occasion recently to run the first version of my astronomy software (star charting, ephemerides, etc.), written for MS-DOS in 1993. (Done on Linux inside DOSBox.) I can't even say that I was under much pressure at the time, and I did the best I could. But looking at some of the decisions I made back then makes me wince.

[Anonymous Coward]

> It's easy to scoff but the industry was a lot younger then and the pressure to deliver these products was sky high.

If by "younger" you mean naive, and if by "industry" you mean Microsoft then yep. There were other software companies at the time were capable of dealing with these kinds of issues perfectly well but Microsoft had a "we know best" chip on its shoulder.

[mhoulden]

I've had the, er, pleasure of using MS Sourcesafe. How bad was Source Library Manager that they considered it to be an improvement?

[BinkyTheMagicPaperclip]

Sourcesafe isn't reliable when performing operations on many files, and worse it won't tell you about failures. It doesn't have atomic commits either. So overall it's an absolute disaster.

SLM was new to me, but if you have a quick search you'll find more detail about it and its brief commercial offshoot Microsoft Delta. One key difference is that SLM can manage large software products, whilst Sourcesafe can't.

Microsoft then moved on to Source Depot which is modified Perforce. Perforce is a solid SCM, and it's still used at work. We've had very few issues with it. (Other projects use Team Foundation or Git, but it would take too much effort to be worth shifting the codebase inside Perforce elsewhere.

[Adam2209]

A case of insensitivity

The irony of the dev teams relying on this while making us a case insensitive filesystem... For anyone who remembers trying to rename file.txt to File.txt

[D-Coder]

Is there anyone reading this website who does not check https://devblogs.microsoft.com/oldnewthing/ on a daily basis???

[Anonymous Coward]

Seemed a bit odd...

sizeof (vector) is the storage the array vector requires in bytes ie the size need for malloc(3) etc

sizeof (vector) / sizeof (vector[0]) is the unitless number of elements in vector.

Would seem that the win95 coders hadn't groked that the two for them were numerically equal was contingent on sizeof (char)==1

When dealing with arrays of other types this blurring could be a rich source of errors of which MS doesn't have a monopoly but does vigorously strives for one.

Even the earliest C compilers would have managed constant folding so premature optimisation was unlikely the culprit.