Comments on: callstack trace on the nintendo ds

By: erisu

erisu — Thu, 14 Jan 2010 13:14:46 +0000

Wow, nice! Pondered this problem a bit myself. ARM/gcc conspire to make it difficult but this solution seems workable. ^^

By: smiker

smiker — Fri, 11 Dec 2009 01:21:52 +0000

Hi, mr Peter Schtraut, i write down here because i dont know where to else.
Im in a serious need of your library in its finished version to be able to make a long time waited tool for chiptune musicians. Im not in a mood of earning money with it, and i have been waiting, but i cant afford it. Its way too expensive and we’re in crisis. I promise i would pay it to you later.
I need HAMLib. My name is smiker, feel free to google me a bit if you wanna know who i am.
Thank you.
Smiker.
It would be…a great christmas gift for an unknown person.

By: Peter Schraut

Peter Schraut — Wed, 02 Sep 2009 08:41:04 +0000

@sylvainulg: From what I’ve seen in the compiler generated assembler code, Stacktrace must be able to branch in both directions in order to find all pop/bx instructions. Take a look at the code below “Support for the unconditional branch was important”, this should clarify.

@Eduardo Costa: It only works with thumb code, because of the devkitARM / libnds makefiles default to thumb mode and this is what I (and many other people) use. The Stacktrace source code is available, feel free to add support for arm instructions.

By: Eduardo Costa

Eduardo Costa — Tue, 01 Sep 2009 20:03:09 +0000

Awesome! Great hack! BTW, why does it only works with thumb mode??

By: sylvainulg

sylvainulg — Mon, 17 Aug 2009 12:43:32 +0000

impressive. I’ve been trying to get something like this from quite a while (and manually doing it as well with the stackdump libnds offers).
A possible tweak to find the end of the function faster, maybe: only branch forward, rather than “Branch only to the target address when it’s different from PC”