Date: Thu, 27 Mar 1997 19:13:45 -0800 From: John Polstra <jdp@polstra.com> To: smc@servtech.com Cc: dyson@freebsd.org, hackers@freebsd.org Subject: Re: Anyone else seen this? Message-ID: <199703280313.TAA28286@austin.polstra.com> In-Reply-To: <333AA089.41C67EA6@servtech.com> References: <199703270427.XAA04344@dyson.iquest.net> <333AA089.41C67EA6@servtech.com>
next in thread | previous in thread | raw e-mail | index | archive | help
In article <333AA089.41C67EA6@servtech.com>,
Shawn Carey <smc@servtech.com> wrote:
> The problem is not GDB related, GDB just makes it clear when it
> happens. When GDB has stopped the program (which is every time so far),
> I have had no breakpoints or watchpoints set.
Hmm, well I've just been messing around with it, and to me it does seem
related to GDB. At least, I can make it happen reliably with GDB,
but not without GDB. Mind you, I don't think it's a GDB bug. It's
just that GDB writes into the text segments of the executables.
On a -2.2 system (unknown vintage, sorry), I compiled hello.c with
"cc -g":
#include <stdio.h>
main(int argc, char *argv[])
{
printf("Hello, world!\n");
return 0;
}
Then I started gdb on it, and set a breakpoint on "main". I ran it
repeatedly (all from the same gdb invocation). Many, but not all,
of the times, GDB gave me the message that the file had changed.
I just tried it on a -current system from March 17th, and it
exhibited the same behavior.
In another window, I did repeated "ls -lT" commands on the
executable. Its modtime did in fact change, though the appearance
of each change lagged a bit behind the moment when GDB noticed the
difference. I am pretty sure that lag was caused by the delays
between "update" runs. (A "sync" reliably made the changes
visible to "ls".)
Hmmm ... Hey, this is pretty interesting. I set the breakpoint at
main and alternamed "r" "c" "r" "c" ... over and over rapidly.
(I.e.: run the program, continue past the breakpoint, program
finishes, repeat.) It looked very much like I got the "has changed"
message from GDB about every 30 seconds, which is the kern.update
interval.
Does this give you something to chew on, John Dyson?
--
John Polstra jdp@polstra.com
John D. Polstra & Co., Inc. Seattle, Washington USA
"Self-knowledge is always bad news." -- John Barth
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199703280313.TAA28286>