From: Groepaz (groepaz_at_gmx.net)
Date: 2003-04-25 00:33:50
On Thursday 24 April 2003 23:03, Ullrich von Bassewitz wrote:
> > i fail to see the problem with that :o)
>
> It violates 2.: Needs extensions and is not easy. How is the linker
> supposed to know that __STACKSIZE__ defines the size of the stack? This is
> what I mean with expressions, a reasonable implementation would need
> something like
>
> MEMORY {
> RAM: start = $7FF, size = $c001, define = yes, file = %O;
> STACK: start = __RAM_START__ + __RAM_SIZE__, size = __STACKSIZE__,
> define = yes, file = "";
> ...
> }
exactly what i said, you insist on doing the dodgy calculation as is, but in
the linker script :) i say its not needed. tell the linker to read the size
from the symbols and all is fine. please explain why you think what i say is
not sufficient coz i really dont get it. except that other linkers work like
you say, what is the reason not to do it like i suggested? ofcoz the linker
can know about __STACKSIZE__ if you tell him, it must read the linkerscript
at first (unless my understanding of the inner workings of the linker is
totally wrong) and then it knows about the values of defined symbols. it can
then link everything, including the stack segment (whose size is known). it
only needs to know that the heap segment gets linked last, since the linker
must know the remaining space in the memory area you load the heap segment
into. (if the heap size is specified in linkerscript, heap can be linked like
any other segment).
however, if you just dont like the idea of having a few things like
"__STACKSIZE__" hardcoded in the linker (your comment suggests that :=)) and
would prefer the generic (imho overkill here) approach you will indeed need
expressions and calculate things in the linkerscript like you say....i'm not
quite sure if that wouldnt be shooting a fly with a cannonball though :=) no
doubt, beeing able to use expressions would add a LOAD of flexibility over my
simple approach...the question is, what would be impossible with the simple
variant and how much would that rare exception justify the lot of work for
the implementation of the expression stuff...i dont quite recall ever having
calculated anything but stack/heap stuff in linkerscripts to be honest (and
i've written a bunch of them for various consoles etc)
oh well :=P maybe i just like kludges in linkercode better than in the library
:=) (this hax0red _heap.s in nes stuff really sux eh? :=P)
gpz
----------------------------------------------------------------------
To unsubscribe from the list send mail to majordomo_at_musoftware.de with
the string "unsubscribe cc65" in the body(!) of the mail.
This archive was generated by hypermail 2.1.3 : 2003-04-25 00:40:30 CEST