[PATCH] Add support for the Win32 hook prologue (try 5)

[Stefan Dösinger]

Am Tuesday 22 September 2009 23:20:12 schrieb Richard Henderson:
> static_chain_on_stack will only ever be true for nested functions.
> You should not need to apply this attribute to a nested function,
> as such functions are never exported.
>
> You could change handle_abi_attribute to reject this attribute
> for nested functions by checking decl_function_context (*node) == NULL.
This doesn't seem to work. I added this:

if (decl_function_context (*node) == NULL)
  {
    error ("ms_hook_prologue is not compatible with nested functions");
  }

in ix86_handle_abi_attribute, but when I try to compile a function with this 
attribute I get this error:

test.c:2:1: internal compiler error: tree check: expected tree that 
contains 'decl minimal' structure, have 'function_type' in 
decl_function_context, at tree.c:7993
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

I am also not sure if ms_hook_prologue is not useful to nested functions. I 
can imagine something like the attached test2.c program. I don't know if it 
is valid C, and I also don't know if the function attribute can be used by 
any other language that has better support for nested functions(in C they're 
a gcc extension as I understand).

I think for now I'll allow the attribute on nested functions, and revert the 
mov and pop like I do if no frame pointer is used. I can imagine that the 
pushl.s %ebp can be used instead of the push the ix86_static_chain_on_stack 
generates for stack adjustment. However, Wine won't ever need this so this 
would be an optimization that won't get much real world test coverage.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: test2.c
Type: text/x-csrc
Size: 455 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20090923/1115b5e2/attachment.bin>