[Mono-devel-list] [PATCH] GetFolderPath improvements

[Todd Berman]

Ben Maurer wrote:
> On Tue, 2004-06-15 at 20:14, Todd Berman wrote:
> 
>>So we are moving a chunk of code from corlib to the runtime to prevent a 
>>  switch in a method that is not used more than once or twice if even at 
>>all in an average program's lifecycle?
> 
> 
> Removing the switch was not the point of this patch.
> 
> My point was that there is alot of data about the system that we gain
> from autoconf, and from the ability to call system apis (such as
> g_get_home_dir)
>      A. must pass to corlib via some internal api (requires adding an
>         icall or
>      B. deal with in unmanaged code
> 
> Both methods require an icall. But method B keeps all the code in one
> place whereas method A requires syncing between corlib and the runtime. 
> 
> Doing this in unmanaged code leaves a nice clean interface.
> 
> Also, note that there are very few places in which we test for Unix
> right now. The only places where I can see this sort of test are in
> Console and System.Security.Principal. OTOH, there are tons of tests
> like this in metadata.
> 
> -- Ben
> 
> 

I don't quite understand why, I don't see any code you are doing now 
that can't be accomplished via Path.Combine 
(Environment.GetEnvironmentVariable ("HOME"), something);

in fact, I know inside MD, we perform at least 2 or 3 of those string 
manipulations outselves (to follow the fd.o standard).

So again, I ask why moving this into unmanaged code is needed.

--Todd