Re: NT fix

[Steven Knight]

On Fri, 20 Oct 2000, Kolarik, Tony wrote:
> This is to fix the problem with hard links that I've been going on about.
> 
> I changed 2.0.2 and ran my version of it through Steven's 3.0 test suite and
> got one fewer failures with it than with the vanilla 2.0.2, although it
> looks like it had nothing to do with my changes. 

The separate cons-test package is actually deprecated.  The tests are
now maintained out of the same source tree as Cons itself, so every
time we release a new Cons, there's a corresponding package of tests
(with the same version number) that's the most up-to-date for that
version of Cons.

Unfortunately, because I had already numbered the separate cons-test
packages up to 3.0, it seems like that would be a more recent copy of
the tests than 2.0.2, but it's not.  I'll add some text to the FAQ and
change the old cons-test web page to clear this up.

> I've attached a diff of 2.0.2 against my version.  

Thanks.  I think this change makes sense.

> The last hunk is there because I get screenfulls of warnings about not being
> able to change modification times. My sources are normally read-only
> versions under source control, so I have always commented it out.  I should
> add a trivial test with a read-only source and a Link()... I can't
> understand how this doesn't bite everyone one else.

This was already fixed in Cons 2.1.2 by chmod'ing the file writable
before changing the modification time, and then re-chmod'ing it to the
actual permissions if necessary.

Since NT doesn't allow the modification time to be changed if you don't
have write permission on the file, this would definitely have been a
problem for anyone else trying to use Link() with read-only source on NT.
I can only assume that there weren't too many people in that situation,
and that they were either ignoring the messages or commenting out the
lines like you did.

        --SK