[Mono-devel-list] reflection.emit in managed code

Jackson Harper jackson at latitudegeo.com
Sun May 11 10:19:42 EDT 2003


Hello,

	PEAPI does support custom attributes now, or at least they are in the
documentation. 

	I am not sure how reliable PEAPI is yet. From my preliminary use it
seems very good, but I have not done that much testing. Perhaps it would
be best to wait a few months untill the ilasm assembler is completly
done and tested before we make a decision. This would allow us to ensure
it is bug free and do some performance testing.

Jackson



On Sun, 2003-05-11 at 06:20, Paolo Molaro wrote:
> On 05/09/03 Varga Zoltan wrote:
> >   I'm thinking about embarking on a project to rewrite parts of 
> > reflection.c in managed code, especially the metadata
> > emission part , 
> > using PEAPI or something like that.
> > Is this a good idea? What do you guys think?
> 
> After I got mcs bootstrapping in mono, many months ago, I though
> about moving part of the reflection.c code to the managed world.
> My plan was to do it after mono 1.0, though. A great deal of
> optimizations and bug fixes went into reflection.c and I'm afraid a
> rewrite at this point would mean losing a lot. Note that some of the
> tweaks are needed to get the Reflection API to output a corlib
> assembly, because the standard R.Emit stuff wouldn't allow for that.
> Also, at this point, I don't think PEAPI is up to the task, last I
> checked it didn't support custom attributes, IIRC.
> I'm not sure how much moving the code to the managed world
> interfers with the ability to emit and execute code at runtime:
> you'd have to change ILGenerator to at least having two backends
> (one for fixups for running the code and one for output to PEAPI).
> This may have a noticeable overhead in the end, that we'd like to avoid.
> Is there a particular reason you want to move the code to the managed
> world? If there is no pressing need, I would hold on on this for the
> next few months.
> 
> Thanks.
> 
> lupus
-- 
Jackson Harper <jackson at latitudegeo.com>




More information about the Mono-devel-list mailing list