Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!lsuc!hcr!larry
From: larry@hcr.UUCP
Newsgroups: comp.unix.wizards
Subject: Re: Stack trace from inside a program . . .
Message-ID: <2925@hcr.UUCP>
Date: Mon, 9-Nov-87 10:04:03 EST
Article-I.D.: hcr.2925
Posted: Mon Nov  9 10:04:03 1987
Date-Received: Tue, 10-Nov-87 02:40:19 EST
References: <151@ncifcrf.UUCP> <7741@steinmetz.steinmetz.UUCP>
Reply-To: larry@hcr.UUCP (Larry Philps)
Organization: HCR Corporation, Toronto
Lines: 27
Keywords: dbx 'where' C innards

In article <7741@steinmetz.steinmetz.UUCP> dawn!stpeters@steinmetz.UUCP (Dick St.Peters) writes:
>In article <151@ncifcrf.UUCP> randy@ncifcrf.UUCP (The Computer Grue) writes:
>>    I have a question for the net: how would one go about writing a C
>>  function that would give you a stack trace of all calls to that
>>  function?
>
>I did this once under VMS.  The experience strongly suggests such a
>function would be machine-, system-, and even compiler-dependent.
>(It was different for the two C compiler I had.)
>
>True in general?

Yup.  It definitely depends on the format of the call frame, which
itself is hardware dependent.  It does not have to be compiler
dependent, but that depends on the compiler writer.  If you want object
files from different compilers to be able to link together, both better
just stick to the machine spec.

Anyway, the easiest way to do this is to go find the source to adb, sdb,
dbx, or whatever on your machine and find the routine that implements
the stack trace.  Steal it and modify it to fit into your machine.
Just put a couple of "asm" instructions at the front to get the stack
pointer and frame pointer registers into some local variable, and you
are set.

If you don't have the source to any of these programs, read the machine
and/or compiler spec, and find out the format of a call frame.  Sorry.