Mail Archives: djgpp/2020/03/17/00:03:58
On Tuesday, August 7, 2018 at 8:38:28 PM UTC-6, rug DOT DOT DOT AT gmail DOT com wrote:
> Hi,
>
> On Saturday, August 4, 2018 at 8:28:27 PM UTC-5, Mdasoh Kyaeppd wrote:
> >
> > mk> meanwhile, I would like to compile pth207s.zip, which has a binary.
> >
> > r.> Also, there's already prebuilt binaries (from years ago):
> > r.>
> > r.> Or did you patch/improve your local copy?
> >
> > no, I have not improved on this code. It isn't compiling properly.
> > yes, I acknowledge(d) there is a binary, it seems you're pointing
> > that out. I hope this thing can make it back into the realm of
> > Free software. thanks, though.
>
> It is still Free, is it not? What host OS are you trying to build atop?
> Are you just worried that it won't rebuild cleanly? A lot of stuff
> indirectly relies (too) heavily on NTVDM, e.g. LFNs.
Having thought for a long time that this package would be impossible to
rebuild, I had tried again today. Swapping out watt32 for libsocket
0.8.0 seems to have done the trick. I now have, not libpthread.a, but
libpthsem.a So I will release my cross-compiled code here:
http://show.ing.me/pth-dj/pth-2.0.7-built.zip
> > r.> Though I vaguely remember that the (2.03p2) /current/ binary was
> > r.> fine but (2.04) /beta/ was accidentally compiled as 686+ only.
> > r.>
> > r.> Why not use Juan's recent build of FSU Pthreads?
> >
> > It seems I might try and compile the source for that now...:
> > ok: http://show.ing.me/pth-dj/fpt0001.tar.gz is compiled.
>
> Okay, but I haven't looked closely myself. Should I? What
> functionality does the prebuilt library lack that you need?
>
> > Let me know if you think we can build the gnu portable threads.
>
> I'm sure someone can, but I don't know what's wrong with just using
> pre-existing binaries. Then again, I understand wanting to be able
> to rebuild various DJGPP things, but it's not always easy.
Glad to have done it today!
> > Also, I am curious why djgpp has not yet implemented job control?
>
> DJGPP is basically only for DOS. Even under Win9x or NTVDM, there were
> no DOS APIs for such a thing.
>
> It's possible that DR-DOS 7.03 had some APIs, but nobody cared enough
> to support it with DJGPP. Nine years ago I did submit a sloppy patch
> for dosexec.c to use int 2Fh, 2707h, but that only worked with DR-DOS'
> TASKMGR loaded. It did work, I think, but I never used it much, plus
> I don't even use DR-DOS anymore. And nobody else seemed to care.
Having tried now to build bash with --enable-job-control it appears
there are a few technical setbacks, yes. I have enumerated the following
signals as missing in DJGPP: SIGCHLD SIGCONT SIGTSTP SIGTTIN SIGTTOU.
Do we have the functionality in DJGPP to perform most of those signals?
Or is this simply a fundamental lapse in OS support that would require
rewrites to basic DOS functionality, as you seem to indicate?
> Honestly, something like DOSEMU is now probably considered a better
> supported Free solution for multitasking DOS stuff.
Hopefully this will change in the future. I bet it will.
- Raw text -