Date: Mon, 16 Sep 1996 15:26:53 -0300
Message-Id: <1.5.4.16.19960916122832.29a7d2c4@dmeasc.rc.ipt.br>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
To: kagel AT dg1 DOT bloomberg DOT com, thoni AT softlab DOT se
From: Cesar Scarpini Rabak <csrabak AT dce03 DOT ipt DOT br>
Subject: Re: Possible static data zeroisation bug in DJGPP v2?
Cc: djgpp AT delorie DOT com

At 08:30 16/09/96 -0400, kagel AT quasar DOT bloomberg DOT com wrote:
>   From: Thomas Nilsson <thoni AT softlab DOT se>
>   Date: Fri, 13 Sep 1996 19:23:43 +0200
>
>   I have an application which I mainly develop on Unix, however thanks to
>   the wonderful DJGPP it's very easy to port to PC's, thanks all who have
>   contributed in the DJGPP project!!
>[SNIP]
>   The structure is declared static at the top level of the module and in
>   the beginning of the file, but still gdb says 'No symbol xxxx in current
>   context'. The structure *should* not be used before that point in the
>   running of the program, but it is not zero as I would have expected.
>
>Static and global variables are not initialized or zero'd unless explicitly
>initialized in the code.  You have just been lucky because UNIX zeros the

Excerpt from K&R Book 2nd ed. pg 40: 
"...External and static variables are initialized to zero by default."

I have no ANSI/ISO standard handy by now, but it is on the standard as well
(I think on section four).

>memory pages it assigns to a process so the garbage you have been getting just
>happened to be the zeros you wanted.  DOS does not zero memory nor does the
>DPMI spec require this (IMHO) so DJGPP does not zero uninitialized memory.
>
>   This being strange, but still I wanted to investigate this structure, so
>   I removed the static declaration to make it more visible, and thought
>   that I might get at it already at the start of the program. And of
>   course I could, *but* now it is zeroised correctly and not destroyed so
>   the program works perfectly...
>
>Just lucky that globals are probably allocated before statics so now your
>structure is on a different memory page with different garbage.
>
>You really should convert the definition of the structure to an initialized
>definition.  Never rely on the value of a variable you have not explicitly
>set.
>
>   Is this a bug, or if not, how do I find out when the structure is
>   trashed. Note however that Purify (a lovely memory access analyser for
>   many plattforms) does not indicate any problems on Unix (so at least
>   it's not a general bug, but then which ones are...)
>
>Purify reports no problem because none exists you are just suffering from GIGO.
>
>   ------------------------------------------------------------------------
>
>-- 
>Art S. Kagel, kagel AT quasar DOT bloomberg DOT com
>
>A proverb is no proverb to you 'till life has illustrated it.  -- John Keats
>
>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Cesar Scarpini Rabak                      E-mail: csrabak AT ipt DOT br
DME/ASC                                   Phone: 55-11-268-35221Ext.350
IPT - Instituto de Pesquisas Tecnologicas Fax:   55-11-268-5996
Av. Prof. Almeida Prado, 532.  Sao Paulo - SP 05508-901 BRAZIL
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~