Nicholas Clark nick at
Fri Feb 8 05:40:46 PST 2002

(sorry for doing a an AOL "me too")

On Fri, Feb 08, 2002 at 01:17:00AM +0100, John Tytgat wrote:

> based on case*in*senstive filename lookups :-(.  And I don't
> believe the "although filing systems can do so" bit as when the
> FileSwitch caching is case*in*sensitive, how can a real FS sitting on
> FileSwitch be casesensitive ? During the development of at least two

Quite. I came to the conclusion that what the PRM said about insensitive
being possible was rubbish because of the reasoning you state. (Although
I've never tried developing filesystems, so don't have the practical
experience of its annoyances)

Even more so when IIRC RISC OS 3.1 added some wildcarding functions to
FileSwitch, and encourage filing system authors to use them rather than
implementing their own code for wildcard matching. "Hangon," I thought,
"these new functions have to be case insensitive to do what most callers
would expect of them, yet you don't provide any flags to them to give
case sensitive wildcard matching for the case sensitive filesystems you
purportedly support".

[By which I mean a case insensitive wildcard matcher would think that "foo*"
should match "foo", "foo1", "Foo2", "fOO3" etc]

But in theory as FileSwitch is a module, and everything is done with vectors,
only time is stopping someone from providing a complete re-implementation
of FileSwitch which can do case sensitivity.

Nicholas Clark

More information about the gcc mailing list