From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <10108100454.AA13597@clio.rice.edu>
Subject: Re: Selector Exhaustion
To: pavenis AT lanet DOT lv
Date: Thu, 9 Aug 2001 23:54:24 -0500 (CDT)
Cc: djgpp-workers AT delorie DOT com
In-Reply-To: <3B72D3CB.11100.FB5F53@localhost> from "pavenis@lanet.lv" at Aug 09, 2001 06:17:47 PM
X-Mailer: ELM [version 2.5 PL2]
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

This is my patch vs cvs:

*** dosexec.ori Wed Aug  1 20:33:04 2001
--- dosexec.c   Thu Aug  9 23:34:58 2001
***************
*** 173,177 ****
    const char *progname;
    unsigned proglen;
!   int i;
    unsigned long fcb1_la, fcb2_la, fname_la;
    int arg_len;
--- 173,177 ----
    const char *progname;
    unsigned proglen;
!   int i, s1, s2;
    unsigned long fcb1_la, fcb2_la, fname_la;
    int arg_len;
***************
*** 379,382 ****
--- 379,383 ----
    r.x.es = parm_la / 16;
    r.x.bx = parm_la & 15;
+   s1 = __dpmi_allocate_ldt_descriptors(1);
    __dpmi_int(0x21, &r);
  #if 0
***************
*** 385,388 ****
--- 386,393 ----
    tbuf_selector = 0;
  #endif
+   s2 = __dpmi_allocate_ldt_descriptors(1);
+   fprintf(stderr,"Dosexec selectors=%x:%x\n",s1,s2);
+   for(i=s2;i>=s1;i-=8)
+     __dpmi_free_ldt_descriptor(i);
    /* Work around the W2K NTVDM bug; see dpmiexcp.c for detailed
       explanations.  */

I built make (and only make) with this change.  On my Windows 2000 system
here is a snapshot of rebuild of libc:

Dosexec selectors=27f:2af
gcc ... -c d0200.S
Dosexec selectors=27f:2af
Load error: no DOS memory
Dosexec selectors=27f:2af
make.exe[2]: *** [d0200.o] Error 1
Dosexec selectors=227:257
make.exe[1]: *** [all_subs] Error 2
Dosexec selectors=1cf:1ff
make.exe: *** [all] Error 2

So, my problem with selectors went away with the above patch, but now I
die due to no DOS memory :-P (it does run longer).  Now what is strange
is before I added the diagnostic print I think I still leaked, but I
can't be 100% sure my make build was correct.  So, does NTVDM leak dos memory
too?