hush - memory leak?

[Rob Landley]

On Wednesday 29 December 2010 13:46:25 Denys Vlasenko wrote:
> On Wednesday 29 December 2010 15:37, Piotr Grudzinski wrote:
> > > Can you compile the following test program:
> > >
> > > #include <sys/types.h>
> > > #include <unistd.h>
> > > int main(int argc, char **argv)
> > > {
> > >    while (1) {
> > >        write(1, ".", 1);
> > >        execv("/proc/self/exe", argv);
> > >    }
> > >    return 0;
> > > }
> > >
> > > via "gcc -Wall testprog.c", run it and report whether you see
> > > free memory disappearing?
> > >
> > > --
> > > vda
> >
> > NO. The memory is NOT disappearing.
>
> Ok, I'm out of ideas.
>
> Since I can't reproduce it on ordinary x86, can you try
> to reproduce the bug in one of Rob's Aboriginal Linux
> prebuilt qemu virtual machines? He has a number of arches
> supported, maybe yours is in that set.

/me goes and reads the thread...

Er, uClibc 0.9.29 on nommu?  From where?  I don't think vanilla 0.9.29 
supported nommu, at a guess this is the uClinux tree?  On what hardware?

I just fired up the armv5l system image under qemu and ran the loop under hush, 
and it's not leaking memory.  The numbers jitter a bit, but only within about 
a 16k range.  I just tried i686, same result.  Powerpc... same result.  
x86_64: same result.  That's both endiannesses, 32 and 64 bit.  Admittedly 
none of those are nommu, I've been meaning to add a nommu target but haven't 
gotten around to it yet.

Still: this is not a busybox bug, this is a bug in your C library that hush is 
merely triggering.

> You can found Aboriginal Linux here:
>
> http://landley.net/aboriginal/

I'm a month overdue for the 1.0.1 release.  I'll get one out later today. :)

> If you will have the bug reproduced in a qemu virtual box, then I
> will be able to do the same, and work on tyhe bug locally.

Rob
-- 
GPLv3: as worthy a successor as The Phantom Menace, as timely as Duke Nukem 
Forever, and as welcome as New Coke.