Avoid duplicate names in /proc/registry (which may crash find) ?
Tue Dec 2 13:38:00 GMT 2008
Corinna Vinschen wrote:
> Some random thoughts:
> - Do we really need d_type support in the registry? How often is that
> actually used?
/proc/registry is probably rarely used, so d_type support is not
But adding it would be trivial and the effect is significant:
Speed up of "oldfind" (which does not crash :-) is >= 2.
> - If find crashes in this situation, isn't this a bug in find which
> should be fixed in find?
Yes. As far as I understand POSIX, it does not require "directory
snapshot" semantics for readdir(). So if e.g. a directory "foo" is
replaced by a file "foo" during a readdir() sequence, duplicate names
with different d_type may occur (at least in theory :-)
> - /proc/registry is a convenience for reading the registry. Due to
> some funny definitions of the registry it's not a full solution. You
> can't write this way, you can't even access the "(Default)" key
Access to (Default) works for me with "/proc/registry/.../key/@"
> I personally don't worry if some border cases don't work. For the
> border cases and for the full access we have regtool.
> Having said that, I really don't know what's the best solution for
> these problems would be. None of the above ideas are very attractive.
> For instance, in idea 1, the path would actually be
3. Add an extension (describing its type) to value only:
But then a possible "somekey.type" should be escaped ...
> Shouldn't we rather defer these border cases to regtool?
Yes, further enhancements to /proc/registry are probably not worth the
Hmm... "/proc/registry/..." pathnames as an alternative for the
KEY/VALUE parameters of regtool would allow to use shell tab-completion
and therefore enhance usability (?).
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Problem reports: http://cygwin.com/problems.html
More information about the Cygwin