[cfe-dev] RFC: A virtual file system for clang

[Manuel Klimek]

On Thu, Feb 13, 2014 at 1:53 PM, Rafael Espíndola <
rafael.espindola at gmail.com> wrote:

> > Current plan is to expose functions to get the VFS description file,
> > which Clang can parse and build the VFS accordingly.
> >
> > Nevertheless, I think we can provide both interfaces, if we freeze the
> > 'Status' data structure in libclang, or if we declare that those
> > functions are exempt from libclang ABI stability.  But in that case
> > one can probably just use the C++ API directly, so I don't see the
> > function-based interface as important.
>
> OK, that is a relief :-)
>
> I was truly afraid we would end up putting part of  the current file
> API on stone because of the libclang abi stability.
>
> I took a quick look at the API last night. An observation is that 99%
> of the users want the simplest version of any api (exists returning
> false on error, create_directory that hides if the directory already
> existed or not, etc). Unfortunately, quiet a few end up using the more
> complex apis.
>
> So what I would suggest for the virtual fs is to start with only the
> simple ones. If you do have a use that needs the more general case,
> adding that predicate with a longer name is probably a good thing so
> that it doesn't get used accidentally.
>

There seem to be 2 interfaces here: one user interface, and one for
subclassing.

Are you fine with the interface for subclassing being the current
Status-based proposal?


>
> Is that OK?
>
> Cheers,
> Rafael
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20140213/67a7cd90/attachment.html>