Re: [MiNT] opendir() regression

[Alan Hourihane <alanh@fairlite.co.uk>]

On Sun, 2010-01-31 at 23:28 +0100, Vincent Rivière wrote:
> Alan Hourihane wrote:
> > This isn't really a regression, but a feature :-)
> 
> Sorry, but it is really difficult to accept that.
> 
> I have the same (old) kernel, the same coreutils sources, and a similar 
> compiler. Only the MiNTLib has changed.
> In the very specific case I have described, the old pwd works, but the new 
> one fails. If this is not a regression in the MiNTLib, I don't know what I 
> could call a regression.

Well, you have to take into account that coreutils is now using new
features of mintlib, it's no longer using the same path. If coreutils
falls back on an error with the new path, and goes back to the old,
that's great, but it is under application control when something fails
like this.

> > Basically your new binaries are taking advantage of a new kernel
> > feature. opendir doesn't exist on older kernels and therefore your 'pwd'
> > binary fails. Older compiled versions of 'pwd' doesn't use opendir.
> 
> You may have missed some information in my report.
> The new pwd works fine in most cases, except when it is run from the root of 
> an hostfs drive.

Ah, o.k. that wasn't clear in your previous post.

If it's just 'pwd' at the root of a hostfs drive then quite possibly
hostfs needs a small tweak to be able to handle opening of directories
which wasn't supported before.

I'll take a look.

Alan.