DOS ain't dead

On 20 May 2021 the Free Pascal team released a new version.

This version is a point update to 3.2.0 and contains bugfixes and updated packages, some of which are high priority. In this case a new target was also backported from trunk. -- There is a list of changes that may break backward compatibility. You can also have a look at the FPC 3.2.2 documentation.

Home page: https://www.freepascal.org/
Download: https://www.freepascal.org/download.html

There doesn't seem to be a release for DOS so far. (Only fpc-3.2.2.i386-win32.cross.i8086-msdos.exe)

---
Forum admin

Yes, you are true. From some reason there is no official 32-bit DOS version for far.
However you can still download the 3.2.2 snapshot here:
ftp://ftp.freepascal.org/pub/fpc/snapshot/v32/i386-go32v2/

I have not tested is so far.
I would be happy if it has fixed a annoying bug 0038470 from FPC 3.2.0 which I reported some time ago (and I also sent a private email to Pierre Mueller /but got no response from him/)
https://bugs.freepascal.org/view.php?id=38470

---
DOS-u-akbar!

> Yes, you are true. From some reason there is no official 32-bit DOS version
> for far.
> However you can still download the 3.2.2 snapshot here:
> ftp://ftp.freepascal.org/pub/fpc/snapshot/v32/i386-go32v2/

Nope. Only 3.2.1 or 3.2.3 here.

---
Forum admin

> On 20 May 2021 the Free Pascal team released a new version.

> There doesn't seem to be a release for DOS so far. (Only
> fpc-3.2.2.i386-win32.cross.i8086-msdos.exe)

True. Pierre hasn't reacted in the last few weeks to use either.

> There doesn't seem to be a release for DOS so far. (Only
> fpc-3.2.2.i386-win32.cross.i8086-msdos.exe)

Go32v2 has been uploaded last weekend:
ftp://ftp.freepascal.org/pub/fpc/dist/3.2.2/i386-go32v2/

> > There doesn't seem to be a release for DOS so far. (Only
> > fpc-3.2.2.i386-win32.cross.i8086-msdos.exe)
>
> Go32v2 has been uploaded last weekend:
> ftp://ftp.freepascal.org/pub/fpc/dist/3.2.2/i386-go32v2/

That's great! Thanks.

---
Forum admin

> That's great! Thanks.

Not so great
The included IDE (in UIDEDOS.ZIP file) is compiled without any debug support.
The download page promisses some alternate IDE builds with debugging and VESA supports ON and OFF but such package is missing.
In the FPC 3.2.0 was in ftp://ftp.freepascal.org/fpc/dist/3.2.0/i386-go32v2/separate/idedos_alternative_builds.zip

But for 3.2.2 is not such file.

---
DOS-u-akbar!

Small bump of this thread to remind jhall that the new release of FPC isn't mirrored to iBiblio yet.

Oh, I haven't even updated my main desktop yet (stuck on 3.2.0). Didn't Laaca say that 3.2.x didn't work on HX due to missing imports? Have newer HX versions from Github not fixed that? Has anyone tested? (I still only have the 3.1.1 "4/22/2018" snapshot installed on my FreeDOS jump drive. It works.)

I also had formerly uploaded Trevor Blight's p5c to iBiblio. (That version outputs C code for GCC and works with DJGPP, e.g. GCC 10.3.) So that newer version (r76) should replace the older one (r67).

Not sure what's going on with Scott Franco's P5 (though I'm still relying on my 2017 GPC/DJGPP 2.05 build of old 1.1 [2010]).

Apparently he has a 1.4 version now, but a very simple and quick check didn't get it working! Dunno. Though FPC 3.2.2 (Go32v2) seems to compile my own simple {$mode iso} [7185] sources without errors. Then again, something about the old P5 1.1 sources compiles okay but won't run ("set" errors). Even GPC can have this problem on certain optimization settings, but for FPC -O2 or -O3 or any change there makes no difference. So the compiled "pcom" can't handle "sets" (unlike GPC) but is much, much faster (2 secs. vs. 9 secs.), oddly.

There's also a P6 out there by Richard Sprague, but I haven't even downloaded nor looked at it at all (yet). I found that through random searching and stumbling again upon Hans Otten.

> Small bump of this thread to remind jhall that the new release of
> FPC isn't mirrored to
> iBiblio
> yet.
>

Or you know, you could email me.

I have mirrored it, I think. Did they change how they packaged things in 3.2.2? In 3.2.0 and previous releases, we had zip files like: doc-chm.zip doc-txt.zip dos320full.zip dos320.zip

For 3.2.2, I could only find fpc-3.2.2.i386-win32.cross.i8086-msdos.exe (installer) which requires fpc-3.2.2.i386-win32.exe (Win32) to work. I'm not a Pascal programmer, so I'm not the right person to test this - so I don't know if I've mirrored the right things.

> I have mirrored it, I think. Did they change how they packaged things in
> 3.2.2? In 3.2.0 and previous releases, we had zip files like: doc-chm.zip
> doc-txt.zip dos320full.zip dos320.zip

dos320.zip is the DOS Go32v2 compiler proper (without sources), so "full" means with sources.

Also see SourceForge for Documentation. (N.B. I'd only mirror TXT, CHM [for IDE], and maybe?? HTML [bloat!].)

> For 3.2.2, I could only find fpc-3.2.2.i386-win32.cross.i8086-msdos.exe
> (installer) which requires fpc-3.2.2.i386-win32.exe (Win32) to work.

Not sure how or why or if that's a hard requirement. The cross-compiler has a built-in OMF librarian and assembler, so it doesn't need any obvious external Win32 utils. That part now has its own subdir: msdos/

Just to reiterate from last time (3.2.0, haven't checked new one yet): "subdirs for separately-built libraries for 8086, 186, 286, 386; GENERICS.*.A libs take up 33 MB".

I don't know if it even works under HX anymore. And I doubt anybody would make a proper .ZIP for DOS users anyways. No idea what's on FPC's FTP since modern web browsers sadly all gave up on that ancient protocol. But maybe I'll try newest HX (v2.19pre8 (Aug. 23)) + ppcross8086.exe (manually .ZIP'd) later this week.

> I also had formerly uploaded Trevor Blight's
> p5c to
> iBiblio.
> (That version outputs C code for GCC and works with DJGPP, e.g. GCC 10.3.)
> So that newer version (r76) should replace the older one (r67).

pascal-p5c-code-r76.zip should replace pascal-p5c-code-r67.zip.

> Just to reiterate from last time (3.2.0, haven't checked new one yet):
> "subdirs for separately-built libraries for 8086, 186, 286, 386;
> GENERICS.*.A libs take up 33 MB".

Since there are six memory models (tiny,small,compact,medium,large,huge) and four cpu targets (8086,186,286,386), that means 24 copies of generics.collections.ppu, which are 31 MB each. So that totals about 748 MB for those files alone. (The related generics.collections.a libraries only total 534 kb.)

NOT complaining! Just noticing.

> I don't know if it even works under HX anymore. And I doubt anybody would
> make a proper .ZIP for DOS users anyways. No idea what's on FPC's FTP since
> modern web browsers sadly all gave up on that ancient protocol. But maybe
> I'll try newest HX
> (v2.19pre8 (Aug.
> 23)) + ppcross8086.exe (manually .ZIP'd) later this week.

I avoided keeping everything but "8086" cpu target, and even then it's 20 MB for all the .a + .ppu files, times six for all the memory models. (Sorry, my RAM drive is "only" 200 MB. Yes, my jump drive has plenty more space nowadays.)

It works! I didn't exhaustively test, but the simple stuff is okay (crt, system).

> There's also a P6 out there by
> Richard Sprague, but I haven't even downloaded nor looked at it at all
> (yet).

Although my code is relatively simple, I did test it on various compilers. So, when trying P6, my conclusion is that it's buggy because various programs that work everywhere else don't work there. (No, I didn't report a bug: I don't know if it's actively maintained.)

Because P5 [sic] is so hard to bootstrap, one guy (five months ago) ported it to Delphi. This does NOT (easily, if at all) compile with Free Pascal, sadly. (I shouldn't be surprised, Delphi is a moving target and very complex.) But his binaries when run atop Windows do show similar behavior to my local build of P5 1.4.

(One particular program of mine works fine in P5 1.1 and everywhere else without errors, but P5 1.4 reports "undefined access error" or whatever unless I use {$q-}. The above Delphi version mimics that behavior. It's possible that it's a bug on my end, but it's not fully clear where or why. Neither FPC nor GPC complain. Anyways, I've "mostly" rewritten that program as it was very naive and sloppy.)