OS_GBPB/readdirm monotonic check

John Tytgat John.Tytgat at aaug.net
Thu Jul 28 13:25:45 PDT 2005

In message <c11879914d.peter at chocky.org>
          Peter Naulls <peter at chocky.org> wrote:

> I think this is perhaps for John, as he made the change.  I recall also
> the discussion started by Graham about the contents of a directory
> changing.
> readdir_r() checks to see if R4 has increased after a call to OS_GBPB -
> this happens to not work over SunFish, and I believe, other filing
> systems.  Indeed, the StrongHelp page says not to make any assumptions
> about this value (apart from when it's -1 of course).   Note also the
> comments at the top (I presume, made by John, about it increasing). Is
> this check really meaningful?   I appreciate that there are a host of
> problems that are trying to be resolved with the handling here.

Long time ago I did some work in dirent.c for the CVS port and the comment
+ hack are mine too.  Can I easily reproduce the issue by calling
readdir_r on a (sub-)directory from SunFish mount ? Would this also be
a problem with raFS ? ;-) I think there are indeed better ways to
implement all this.

John Tytgat, in his comfy chair at home                                 BASS
John.Tytgat at aaug.net                             ARM powered, RISC OS driven

More information about the gcc mailing list