Re: [MiNT] opendir() regression

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

On Sun, 2010-01-31 at 14:45 +0100, Vincent Rivière wrote:
> Hello.
> 
> In order to check GCC is OK, I have just recompiled coreutils-6.9. The pwd 
> command does not work as expected in a specific case, probably due to a 
> regression in the MiNTLib.
> 
> Basically:
> root@easymint:/f>pwd
> ./pwd: cannot open directory `..': Internal error
> 
> The kernel:
> The problem happens only with old kernels and Ozk snapshot,
> but it works fine with the trunk and Helmut snapshot.
> 
> The MiNTLib:
> My old pwd compiled from the same sources with an old CVS MiNTLib works fine 
> on every kernel. This supposes that the bug is in the MiNTLib.
> 
> The current directory and file system:
> This problems appears only on ARAnyM's hostfs mounts, when the current 
> directory is a hostfs root, like /f.
> 
> The problem occurs because pwd does always opendir("..").
> As a summary, this fails on old and Ozk kernels when the current directory 
> is an hostfs root.
> 
> This has probably some relation with Alan's evolutions on directories.
> 
> See the attached testcase.

Vincent,

I've checked mintlib for the opendir() call and it hasn't changed since
the previous mintlib release.

Also, even the kernel has only had a one line change to the system call
and that wouldn't interfere.

What is the actual "errno" value you get from your testcase with hostfs
and aranym with the old kernel?

Alan.