X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f From: Rugxulo Newsgroups: comp.os.msdos.djgpp Subject: Re: compiler error on Windows 2003 Date: Mon, 6 Apr 2009 07:16:52 -0700 (PDT) Organization: http://groups.google.com Lines: 61 Message-ID: <20698405-6eed-4a98-9b84-0db7b2712335@k41g2000yqh.googlegroups.com> References:

<16d44b1b-8acc-48b3-bf49-45b71e423d78 AT r37g2000yqn DOT googlegroups DOT com> <2ad157aa-66e6-45eb-9f2c-3816fdbecd58 AT c36g2000yqn DOT googlegroups DOT com> NNTP-Posting-Host: 65.13.115.246 Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable X-Trace: posting.google.com 1239027413 1400 127.0.0.1 (6 Apr 2009 14:16:53 GMT) X-Complaints-To: groups-abuse AT google DOT com NNTP-Posting-Date: Mon, 6 Apr 2009 14:16:53 +0000 (UTC) Complaints-To: groups-abuse AT google DOT com Injection-Info: k41g2000yqh.googlegroups.com; posting-host=65.13.115.246; posting-account=p5rsXQoAAAB8KPnVlgg9E_vlm2dvVhfO User-Agent: G2/1.0 X-HTTP-UserAgent: Opera/9.64 (Windows NT 6.0; U; en) Presto/2.1.1,gzip(gfe),gzip(gfe) To: djgpp AT delorie DOT com DJ-Gateway: from newsgroup comp.os.msdos.djgpp Reply-To: djgpp AT delorie DOT com Hi, On Apr 6, 4:55=A0am, "Tomasz Zbro=BFek" wrote: > Rugxulo wrote in > > > EDIT: Ah, I forgot about this, maybe this will help most of all > > (although I'm not sure those values are truly optimal, I think 0 and > > 4096 will work even better but much slower)? > > > --param ggc-min-expand=3D10 --param ggc-min-heapsize=3D8192 > > hi, > I used these two parameters (10 and 8192 values) with gpp 3.4.4 and final= ly > I compiled and linked our application on Windows 2003. Good! DJGPP gets a reprieve from the governor! (Who needs a bugfix when you can have a workaround?) ;-) > Anyway another > problem is to debug it on Win2003 :))) and I think it won't be possible > with 32MB RAM restriction... our application exe file with debug > information has 19 MB size, I've been told that debugging info isn't loaded in normal execution. > RHide crashes with "divide by zero" error when > I load project file (on XP when project file is loaded ntvdm process take= s > 49 MB) and RHGDB quits with no information when I run the application (on > XP ntvdm process takes 29 MB when application is being run and stopped on > the first breakpoint)... of course on the XP there is no problem to debug > this application with RHide (the newest I know is 1.5c with gdb 6.3) or > RHGDB. > > Again, thank you very much for help Rugxulo! I'm not done yet! :-) You can always use DOSEMU (which has static compiles of it plus FreeDOS on its SourceForge page). It works quite well in my limited testing (although I didn't try debugging anything). Of course, it comes with default 20 MB DPMI limit, but that can be configured (.dosemurc or "dpmi -m 0xA000" will double it, for instance). I tried DOSEMU the other day in CDlinux, and it ran quite well (GdM's 3d engine, Tran's Timeless demo, compiling stuff via DJGPP). Honestly though, CDlinux comes with DOSBox and WINE but not DOSEMU by default (which is weird since DOSEMU is much faster and quite good). But two simple .tar.gz files from the homepage aren't exactly a huge download, so it's easy to install. Also, dare I suggest, but maybe you can use OpenWatcom if your app isn't too DJGPP or GCC specific. It at least supports various non-DOS hosts, including Windows, which can cross-compile for DOS. That won't let you debug, but it might prove useful as an alternative. P.S. Just to be exhaustive for future reference (and you'll have to ask DJ or Eli or CWS for more details), I think you can decrease the transfer buffer and/or default stack size (via stubedit) to save even more memory still albeit not as much as the garbage collection --param values mentioned above.