Subj : Compilers/systems
To   : Vitaliy Aksyonov
From : andrew clarke
Date : Wed Feb 08 2023 06:52 am

On 2023-02-07 07:22:08, Vitaliy Aksyonov (1:104/117) wrote to andrew clarke:

 ac>> From memory the very recent versions of MSVC no longer produce
 ac>> binaries that will run under Windows XP. I've found a good "middle
 ac>> ground" is something like Visual Studio 2012, which still runs well
 ac>> under Windows 10 & 11.

 VA> Visual Studio 2012 is not available anymore. Oldest one is 2013. Would
 VA> it produce binaries for Windows XP?

I'm not sure, but if your C++ code builds with a modern version of MSVC then it will probably build with VS2012, so anyone with that compiler should still be able to build a version that will run in XP unless you're using a new C++ feature from C++11 or C++20 that VS2012 doesn't support.

But for GoldED it shouldn't really be necessary to refactor the code using C++'s increasingly estoric features. Instead, just using features from the STL would be a big improvement.

 ac>> OTOH it would be preferable if a free compiler could be used in
 ac>> Windows. Recently I learned that in 2020 Embarcadero released a
 ac>> fork of Dev-C++ that provides GCC 9.2 and supports C++11:

 VA> MS provides free VS community edition. I don't see any reasons why can't
 VA> it be used.

I'd forgotten about the community edition.

I suppose you could use that, though the modern versions of VS are kind of scary. I mean a 10+ GB download just to compile "Hello world" in C is a bit excessive. The download is bigger than any regular Linux distro. Just for the IDE and C/C++ compiler.

It's strange Microsoft never released a free version of MSVC with just the compiler, header files and libraries but without the IDE. Borland did that in 2000, 23 years ago! I think it even had STL support.

 ac>> I was going to suggest DOS/DPMI support should be dropped, but I
 ac>> see GCC 10.2.0 was ported to DJGPP in 2020, so that's probably
 ac>> useable. Though I doubt many people would complain if DOS support
 ac>> was removed from future GoldED versions.

 VA> That's a good question. Main concern here is that FidoNet is mostly
 VA> retro hobby and people may want to run it on old computers and old OSes.

Well they can still run the older retro versions.

The question really comes down to whether you and other devs are hamstrung by having to still support DOS. Of course if it turns out to be easy to keep supporting it, eg. by cross-compiling with OpenWatcom, then that's good news.

 ac>> Another option for Windows, OS/2 & DOS may be to build with a
 ac>> recent version of OpenWatcom 2.0, though I don't know how well it
 ac>> supports STL or C++11. It may be good enough. The great thing about
 ac>> OpenWatcom 2.0 is can run under Linux, and it's also a
 ac>> cross-compiler, so you can build DOS, Windows & OS/2 apps from
 ac>> Linux.

 VA> I like the idea of using cross-compiling. In this case there is no need
 VA> to setup many different systems. Even if they work on virtual machine.

It does simplify things. It's also really fast.

Though another option for cross-compiling (Linux to Windows, and also MacOS to Windows) is mingw-w64.

 VA> From what I've read about OpenWatcom - they don't really conform fully
 VA> to even C++98 and I'm not sure about C++11 support either. Need to try
 VA> that.

 VA> Even refactor code to use C++98 would be a huge improvement. It's full
 VA> of old pure C approaches.

Yep, modern C/C++ in OW2.0 is a bit hit-and-miss, though there is at least some STL support available, ie. strings, vectors, stacks, etc. It's worth experimenting with it just to learn what its limits are, with respect to which C++ features you'd like to use. It may be good enough.

--- GoldED+/BSD 1.1.5-b20230205
 * Origin: Blizzard of Ozz, Melbourne, Victoria, Australia (3:633/267)

.