Re: [cp-patches] FYI: java-nio adjustments for target/native

[Roman Kennke]

Hi Tromey,

Am Mittwoch, den 18.01.2006, 14:29 -0700 schrieb Tom Tromey:
> >>>>> "Roman" == Roman Kennke <address@hidden> writes:
> 
> Roman> Then we can use shorter macro names or even better function
> Roman> calls like cp_mem_alloc() or something similar.
> 
> Can we use the standard names, and then require the more bizarre
> ports to do things like "#define malloc BlahBlah"?

That would most likely cause problems on these ports. The specific
problem is that these ports actually have a malloc() syscall, only that
it doesn't really do what you would expect, and the macro works around
that by calling another (platform specific) syscall that does what's
expected. The best thing I can offer is a shorter macro name.

> The issue that came up here yesterday on irc is that, apparently, some
> of these macros are only needed for ports that won't be in Classpath.

I am aware of that problem. Still I don't know how to solve that
problem. We simply cannot give away some of our ports, due to license
and other restrictions.

> Roman> -  memcpy (_dst + dst_offset, src, dst_len);
> Roman> +  TARGET_NATIVE_MEMORY_FAST_COPY(src + index,_dst + 
> dst_offset,dst_len);
> 
> There are still platforms without memcpy?  What are they?
> This seems like a great place for a target port to provide a
> replacement function or macro.

I can't quickly find a port that overrides this. Maybe this macro can be
removed again. Let me investigate that.

Roman

signature.asc
Description: Dies ist ein digital signierter Nachrichtenteil