Mail Archives: djgpp/2009/03/13/22:00:06
Hi,
On Mar 13, 4:59=A0pm, Gautier <gaut DOT DOT DOT AT fakeaddress DOT nil> wrote:
>
> >> Of course, I've heard that the Ada front-end compiles itself, so
> >> that's better than nothing, right?
>
> > It is so. I do not however use Ada at all, so I have not done any other
> > tests myself. Perhaps somebody could help.
>
> It's done - of course my biggest hurdle was to find the right system to
> compile on. DOSBox and Vista gave me some compiler crashes (not enough
> memory); XP was perfect.
Although I doubt XP is perfect by any means, I guess here it's less
crappy. I think you'll appreciate the recent quote from AuroraUX:
"Ada was chosen over other languages because it sucks the
least." ;-)
The problem is most likely bad hardcoded default DPMI memory limits
(which Vista, DOSBox, and DOSEMU all share). All three can usually be
overcome with a few settings adjustments (registry,
dosbox.conf, .dosemurc).
> On the results side, there was a strange effect
> on displayed colours, but once all new warnings were gone (mostly about
> assembler insertions that required a "volatile" flag), the problem was
> gone as well.
Confirmed.
> The new GNAT validity checks found three potential bugs.
> I'll send the latest archive to Rugxulo; maybe we'll agree on a
> permanent place for the file ?
I dunno, DJ still has somewhat outdated files of various things
(libkb, lzop, cwsdpmi) although I know of no good way for him to (only
partially) update RSXNT without some license clash. But I haven't had
time to bother learning the (very specific) .ZIP layout just to upload
these to him anyways. I guess anybody who really cares can find the
files themselves.
For now, anybody can get the latest DOS 3D engine here (w/ srcs) :
http://rugxulo.googlepages.com/eng3d018.zip (920k)
> The compilation runs "out of the (dos)
> box" :-). No need to know Ada, just start one of the "make*.bat".
> The demo is probably a good stress-test: some large data put as sources,
> low level things like keyboard & sound drivers, data streaming,
> decompression, etc.
It's definitely way more interesting than most tests. :-)
> Only tasking is untested there. There are classical
> examples (the dining philosophers ?) that could be fine to be tested.
> Not sure what a "serious" test for tasking is...
Me either, but here's a quick link:
http://en.wikibooks.org/wiki/Ada_Programming/Tasking
> > There is however more problems expected with upcoming gcc-4.4.
> > Some time ago I tried to build weekly snapshots of it as
> > Linux to DJGPP cross-compiler and run into trouble with Ada.
>
> > Standard DJGPP does not have sockets, so I had to turn off socket suppo=
rt.
Can WATT-32 not be used?
> > Ada library did not compile even after that. I'm not sure that I'll hav=
e
> > enough time to fix that.
>
> Phew, looks difficult. Anyway, no need to worry too much: I guess that
> the intersection DOS and Ada should not be that big...
Which is a bit ironic considering DOS always being touted as suitable
for embedded systems. (Just don't use VFAT, ugh.)
- Raw text -