From: Nix <nix_at_esperi.demon.co.uk>
Date: 13 Mar 2003 00:22:17 +0000

On 12 Mar 2003, nix_at_esperi.demon.co.uk stipulated:
> Built with -Os, however, the rockbox firmware crashes somewhere between
> show_logo() and browse_root(). (I've not tested it with -O2 yet.)
>
> I'l be debugging this shortly, as -Os yields a 137684-byte ajz file, and
> I want that 5K saving :) I'll try to crash it in the simulator, and if
> that doesn't work

(It doesn't.)

> then I'll have to try moving show_logo() around or
> writing suitably named files to the disk to see where it stops.

sleep() is crashing (the first time it's called, during serial setup; a
one-tick sleep).

Investigation (so far) reveals that flow is constantly entering sleep(),
shooting through it and out in a tick (as it should do), and coming
straight back in at the top again. There's no sign of control ever
returning to the caller (serial_setup()). Bizarre.

(Debugging continues, hampered by needing to pull the power before
Rockbox will boot, and by its insistence on turning off Rockbox whenever
USB fires up.)



Some mindlessly picky nits: apps/alarm_menu.c and
apps/recorder/recording.c shouldn't have their executable bits on.
<http://rockbox.haxx.se/mods/reassemble.html> talks about `bumbers' in
the title. :)

-- 
#ifdef USE_ISPTS_FLAG
		} else {	/* else pty, not pts */
#endif

Received on 2003-03-13