darcs patch: Get External Core (-fext-core) working with
simonmarhaskell at gmail.com
Mon Nov 13 04:46:08 EST 2006
Samuel Bronson wrote:
> On 11/6/06, Simon Marlow <simonmarhaskell at gmail.com> wrote:
>> Samuel Bronson wrote:
>> > Unfortunately I have no idea where to look for the *time*... I wish
>> > non-exported ("static") names were qualified with the module name in
>> > the object files, just like exported names are... (would make it
>> > easier to compare output between cachegrind runs, IMO)
>> I worry slightly about bloating the symbol table and increasing link
>> times, but
>> perhaps I shouldn't. You can make this change pretty easily - see the
>> Outputable instance for Name in the Name module.
> Actually, that won't work without adding the Module to Internal Names,
> will it? Or is there a way to find the module from it somehow?
Yes, that's a fine point. I take it back - it's not very easy to do this after all.
> shouldn't it be possible to strip out the "static" symbols if desired,
> since they aren't really needed for anything?
I don't think this affects stripping out the local symbols. Whether a symbol is
local or not is a property of the symbol, not its name.
> Or even to make the
> inclusion of the module name optional? (I suppose it would need to be
> a static option though.)
Yes, it would be better if it was optional.
More information about the Cvs-ghc