Fellows,
this is actually a proposal, yet code committed:
binutils-current should be updated shortly,
gcc-3.4 is in CVS on SF.
So, this is all about how mspgcc will handle startup for different languages.
Mainly this has been done cause of c++/java/fortran implementation.
This does not discard old approach and backward compatible.
This extends current .text section layout only.
Section .text is broken down into a couple of sections which will be arranged
on a link stage (you will see a few tricks from AVR port):
.init
.init0
...
.init9
constructors list
destructors list
.text
.fini9
...
.fini0
.fini
Subsection '.init' remains a first field of text section. An old style
(gcc-3.2 and earlier) start-up code will reside here with no changes (see
doc.txt).
Subsection init0:
the _reset_vector__ is defined here. only one machine instruction will
be
executed here:
mov #23168, &288 ; disable watchdog
Subsection init2:
in this section all data from data section being initialized.
Subsection init4:
in this section all data from bss section being zeroed
Subsection init6:
being linked only when constructors necesary.
all global constructors will be called here.
stack pointer will be initialized here with RAM top.
so, main's attribute reserve(x) will be useless.
Subsection init9:
jump to main.
constructors list:
data pointers to the global constructors. Being accessed from
subsection
init6 by __do_global_ctors()
destructors list:
data pointers to the global constructors. Being accessed from
subsection
fini6 by __do_global_dtors()
.text:
user code will reside here (including ROMs user defined sections)
Subsection fini9:
this one where __stop_progExec__ label resides.
executes one machine insn:
push r15
Subsection fini6:
being linked only when destructors necesary.
all global destructors will be called here.
subsection fini0:
where real stop prog exec being executed:
pop r15
__stop_progLoop:
bis r15, r2
jmp __stop_progLoop
subsection fini:
for backward compatibility...
User can put his code to sections init1,3,5,7,8, text, text.*,
his.own.sections, fini8,7,5,4,3,2,1
There are no branches between section. All the code in there will be executed
consequentally.
Actually, this might be much easier (better?) to call constructors in main's
code just before first main's instruction after stack initialization, but
then user cannot pass parametres to main().
Actually, this might be much easier (better?) to call destructors just before
returning from main(). But then destructors will not be called when execution
aborts with abort().
So, any suggestion?
cheers,
~d
--
/********************************************************************
("`-''-/").___..--''"`-._ (\ Dimmy the Wild UA1ACZ
`6_ 6 ) `-. ( ).`-.__.`) Enterprise Information Sys
(_Y_.)' ._ ) `._ `. ``-..-' Nevsky prospekt, 20 / 44
_..`--'_..-_/ /--'_.' ,' Saint Petersburg, Russia
(il),-'' (li),' ((!.-' +7 (812) 3468202, 5585314
********************************************************************/
