Supporting Windows from 95 to 10

[Bart]

In the end it just wasn't worth it anymore and now we almost completely stripped all Win9x code (AnsiString API calls) from Lazarus and rely fully on WideString API.
This inevitably will fail in Win9x.

Some of that is mitigated by installing the Microsoft Layer for Unicode on Windows 95, 98, and Me (MSLU) library.

You miss the point.
It was just too much of a hassle to maintain the win9x code and it made developing new features difficult, because none of the developers would even notice that they had broken the win9x part.
E.g. some of the code that interfaces with the debugger only works on NT based OS. So you already need a define to build a partially crippled IDE for win9x.
Notice that I was (and still am) the only developer that has a win9x machine. I used to have a native WinME machine, but that one does not exist anymore, so I have a Win98SE virtual machine now, which will crash randomly. Even installing the compiler on that is a PITA since the fpc installer is NT based. Once we move to Lazarus 1.8 I will kill that VM, because at that time all support for win9x is dropped (see "Phasing out support for Windows 98 and Windows ME").

If there would be a community that would actively contribute to a win9x widgetset, then it could be done, but the demand for this seems to be negligible.
We should however IMO not start mixing win9x code into the current win32 widgetset again.

Bart

[marcov]

Some of that is mitigated by installing the Microsoft Layer for Unicode on Windows 95, 98, and Me (MSLU) library.

Often suggested, few practical reports. A lot is possible, nothing was done, the problem is a manpower one, not a technical one, as Bart already said.

A lot of people like to cross off that item (since saying no to a customer or user is always hard), but nobody wants to work (and work continously) to make that happen and keep that (win9x) option open.

Note this managerial defect has been that way for the better part of 10-12 years, iow since Windows XP came out.

Enough is enough.

[Thaddy]

Note this managerial defect has been that way for the better part of 10-12 years, iow since Windows XP came out.
Enough is enough.

I had to re-read it to understand that Bart wasn't the managerial defect... 
But I think almost everybody agrees with you and Bart.
And there are pointers in place on which versions to use when museum pieces are required to be maintained.

[Bart]

I had to re-read it to understand that Bart wasn't the managerial defect... 

ROFL

Bart

[marcov]

I had to re-read it to understand that Bart wasn't the managerial defect... 

No, and neither were Tomas and Pierre who also have done a lot for win9x (and dos) over the years.

But again, all those efforts were usually around release time. Nobody tested trunk with win9x. It was not continuous.

[palacs]

What about the NativeNT target via cross compilation?

As far as I get it from the wiki page, it could even run on NT 3.51, which means it would be fine on Windows 95 as well.

Anyone tried it?

[Thaddy]

As far as I get it from the wiki page, it could even run on NT 3.51, which means it would be fine on Windows 95 as well.

No. It's is not fine. Even 3.51 has a complete 32/64 bit kernel and sub systems. Win95 hasn't. It is based on win32s, which is a bold on to 3.11.

[BobS]

It's tough to let those old ones go, what memories they bring, but you can see why it's not really worth the effort: https://en.wikipedia.org/wiki/Usage_share_of_operating_systems#/media/File:Operatingsystem_market_share.svg

Even good old XP is at 1.25% at the end of 2015 though it beats out Vista (which I liked).  Now not being able to read German I'm not sure how those stats were collected maybe based on what people were using when they hit a certain web site and so it probably underreports some of those old OSes a bit like the Win95 setups running machinery (and I've seen reports on CNC machines that still use/are stuck on Win95) that never use the internet, but then those people aren't too likely to want to use Lazarus anyway .

[PatBayford]

I find it hard to believe that anyone is using Windows 9x anymore, for anything. What could the possible use case be?

Believe it or not, one of the major problems here is the success Microsoft had selling "embedded" versions of Windows, which can be found in equipment from cheap consumer items, up to million dollar plus medical scanners. These nearly all use ROM based operating systems, so upgrading the OS is all but impossible. Another big hand for the Redmond giant.

[PascalDragon]

What about the NativeNT target via cross compilation?

As far as I get it from the wiki page, it could even run on NT 3.51, which means it would be fine on Windows 95 as well.

Anyone tried it?

The NativeNT target isn't for running Win32 applications, but applications directly above the NT kernel (e.g. the Windows subsystem itself is such an application, or the POSIX subsystem or the chkdsk variant that runs at startup) or drivers loaded by the kernel. So while it's true that applications developed for the target could potentially run on NT 3.51 (I've never tested it though) it's not helping you any further as they're a completely different kind of beast.

And those that want a small Windows system: why not give ReactOS a try? Considering it's based on the Windows Server 2003 API it's still supported by FPC.

[marcov]

And those that want a small Windows system: why not give ReactOS a try? Considering it's based on the Windows Server 2003 API it's still supported by FPC.

Hardware, drivers probably.