Date: Mon, 15 Jul 1996 13:33:08 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: Oberhumer Markus <k3040e4 AT c210 DOT edvz DOT uni-linz DOT ac DOT at>
Cc: djgpp-workers <djgpp-workers AT delorie DOT com>
Subject: Re: gdb crashes if environment too big
In-Reply-To: <199607141656.SAA27063@c210.edvz.uni-linz.ac.at>
Message-Id: <Pine.SUN.3.91.960715132750.13725D-100000@is>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


On Sun, 14 Jul 1996, Oberhumer Markus wrote:

> Yes, you are right. It works fine with stubbed executables.
> The size of the environment is computed anyway, so the
> bug in v2load.c should be easy to fix.

Indeed.  On second thought, it probably is just a typo: the source uses 
si.minkeep instead of si.env_size, but I also took the opportunity to fix 
another suspicious constant and bump up the transfer buffer size.   Here:

*** src/debug/common/v2load.c~0	Thu Jan 11 04:45:10 1996
--- src/debug/common/v2load.c	Fri Jul 12 20:36:18 1996
*************** int v2loadimage(const char *program, con
*** 60,66 ****
    unsigned client_cs, client_ds, my_ds;
    __dpmi_meminfo memblock;
    unsigned new_env_selector;
!   char true_name[200];
  
    _truename(program, true_name);
  
--- 60,66 ----
    unsigned client_cs, client_ds, my_ds;
    __dpmi_meminfo memblock;
    unsigned new_env_selector;
!   char true_name[FILENAME_MAX];
  
    _truename(program, true_name);
  
*************** int v2loadimage(const char *program, con
*** 88,94 ****
      strcpy(si.magic, "go32stub, V 2.00");
      si.size = 0x44;
      si.minstack = 0x40000;
!     si.minkeep = 4096;			/* transfer buffer size */
      memset(&si.basename, 0, 24);	/* Asciiz strings */
    }
    if (header[0] != 0x014c) {		/* COFF? */
--- 88,94 ----
      strcpy(si.magic, "go32stub, V 2.00");
      si.size = 0x44;
      si.minstack = 0x40000;
!     si.minkeep = 16384;			/* transfer buffer size */
      memset(&si.basename, 0, 24);	/* Asciiz strings */
    }
    if (header[0] != 0x014c) {		/* COFF? */
*************** int v2loadimage(const char *program, con
*** 134,140 ****
    si.env_size += 4 + strlen(true_name);
  
    /* Allocate the dos memory for the environment and command line. */
!   i = __dpmi_allocate_dos_memory((si.minkeep + 256) / 16, (int *)&new_env_selector);
    if(i == -1)
      return -1;
  
--- 134,140 ----
    si.env_size += 4 + strlen(true_name);
  
    /* Allocate the dos memory for the environment and command line. */
!   i = __dpmi_allocate_dos_memory((si.env_size + 256) / 16, (int *)&new_env_selector);
    if(i == -1)
      return -1;
  

> BTW, did your recent patch for dosexec.c include a test for
> a possible environment overflow ? Looks like we should add a 
> test for talloc().

Yes, I thought about this; I'll submit the necessary changes soon.