Re: How do I get a .s file assembled into libgcj?

[David Daney <ddaney at avtrex dot com>]

Tom Tromey wrote:

> > > > > > "David" == David Daney <ddaney@avtrex.com> writes:
> > > > > >            
>
> David> The problem I have is that my current solution requires an assembly
> David> language source file and I cannot figure out how to adjust
> David> configure.in and/or Makefile.am for libjava so that the .s file is
> David> compiled from the source tree and then added to libgcj.{so|a}.
>
> It should be as easy as adding the .s file to the appropriate
> _SOURCES macro in Makefile.am, and re-running automake.
>  
Thanks, I am slowly figuring it out.  Conditionally including the .s 
file is now solved.

> However, I think we'd really prefer something more along the lines of
> the other existing ports, if that is possible.  Andrew is the expert
> on this part.  But see how the other ports do it.  My understanding is
> that this is done by writing some glue code (see, e.g.,
> libjava/include/powerpc-signal.h) and also implementing
> MD_FALLBACK_FRAME_STATE_FOR in gcc.
This is that approach I took.  However, currently I still need an 
additional .s file.

> If there's no other way, adding a conditional assembly source would
> be fine.
>
> David> As an alternative I could put the code in a C++ function using the
> David> asm() statement, but I would need some way to force the C++ compiler
> David> to always generate an identical stack frame layout no matter what
> David> optimization or code generation options were used.
>
> Again, Andrew is the expert here.
>
>
> If you're interested in a full MIPS port, you could also implement the
> libffi closure API and the locking code to enable hash
> synchronization.  The former is useful for the interpreter and
> reflection; the latter is more generally useful.
One thing at a time.  I suspect I will be looking into these soon.

Thanks again.
David.