[avr-gcc-list] Ways to fix gcc bug - builtin_return_address

[Andy H]

Anatoly and others.

Well the next bug on list of testsuite execute/torture failures for AVR 
is for gcc's builtin_return_address.

This function is rarely used. But instead of ignoring it, it would be 
better to fix it, that change test to exclude AVR.

Target is supposed to tell gcc the RTL address of the return address on 
the stack - from  frame level. But at present it just adds 1 to frame 
pointer. So its wrong if anything is saved on stack - and/or there are 
local variables on stack) - duh!

We don't need any backs traces and there are other targets that only 
give return address for current frame.  So only level 0 matters. The 
correct way is to get  frame size and subtract that from frame pointer. 
Frame size is only known in late stages of compilation - so there needs 
to be a forward reference.

There would appear to be 3  ways I can do this

1) Create UNSPEC instruction that loads register with frame size (which 
will be known when instructions is output)
2) Create UNSPEC instruction that defines symbol= frame_size.
3) Find some way to get gcc output (2) rather than use UNSPEC.

What do you suggest ?


Andy