Chris Liechti wrote:
3)constructors/destructors jf staitc not called from startup code (as in
previous version, needs libmspgcc.S fixes).
ok.. some knowledgeable person ready to submit patches?
I'll try, but somebody have to check them before adding to oficial
version :)
4) --fdata-sections generates internal compiler error. I think it must
generate message that --fdata-sections not supported.
or the function should work, maybe it could be useful for the
devices >64kB Flash? (putting data in sections above 64k?)
As I saw in sources, this function not implemented for all targets in
gcc 3.4.3. I think, it's better to spend time porting mspgcc to newer
version of gcc.
5) uninstall does not delete root folder (c:/mspgcc).
that is because it is not empty. if it is a file that was installed it's
a bug, but it's common that the examples folder is left over as the
compilation generates files when a user tries them out.
No, noone file remains in directory, just empty directory.
6)--ffunction-sections with --gc-sections causes all code to be removed
from output file. Needs corrections of linker scripts:
.vectors :
{
PROVIDE (__vectors_start = .) ;
*(.vectors*)
KEEP(*(.vectors*))
also, KEEP() must be added to all .initX, .finiX, .ctors and .dtors
sections.
ok, looks like an easy thing, there are just so many files... ;-)
Maybe some (sed) script can help? Unfortunately I have no experience in
*nix text processing utilites.
would there be a similar keyword to ensure that an empty bss section is
also included in the binaries? it seems that GDB can not load files
where no global variables are declared.
I don't know. If GDB fails, the right way is to fix GDB itself, rather
than applying workarounds?
I think I can make a patch to fix linker scripts if required.
that would be great. while i think i understand your changes, i'm not
having much practice working with these files.
ok.
Sergey.
