FreeTDS Development Group

["Craig A. Berry" <craigberry AT mac.com>]

On Mar 22, 2009, at 6:41 PM, James K. Lowden wrote:

> I'm also attaching fwiw the BSD makefile that I use to generate the
> snapshots.  If you're so inclined, maybe you can see why the upload  
> fails
> to happen.
>
> Until a few months ago, the nightly snapshot was handled by a simple
> script that ran every night, provided the system was up.  Then I had  
> the
> bright idea to provide a "nightly" snapshot of the release, too, so  
> that
> people would have immediate access to patches.  But it didn't make  
> sense
> to generate the release tarball nightly, because patches are applied
> irregularly.  I wanted the tarball date to reflect the latest patch,  
> so
> that people could gauge whether or not an earlier tarball they had was
> still the most recent.
>
> How to do that?  Comparing MD5s seemed like the right way.  And still
> does.  But I seem to have been too clever by half, and can't get the  
> thing
> to go forward when it should.  Bah.

I got make dist working on OS X, patch attached below.  There are a  
couple of directories that didn't get created and thus couldn't be  
copied into unless they were included in lists of things to be  
distributed.  I don't actually see how make dist could work anywhere  
unless cp means something rather different on Linux than in BSDish  
places.  So that gets me to where I can, as a last resort, generate my  
own snapshots.

On md5 and snapshots and James's snapshot script, it looks like the  
script calculates the md5 on a directory:

        /usr/bin/md5 $(WORKDIR) > $@

When I try that, I get:

% md5 freetds
md5: freetds: Is a directory

So where I live, you can't do the md5 of a directory.  Where James  
lives, you can, but what does it mean?  One of the things it might  
mean is that on James's system the md5 is doing its magic on the  
directory file itself, not the contents of the directory.  If  
something in a sub-subdirectory changes, I wouldn't think the  
directory file for the grandparent directory would change, so its md5  
wouldn't change.

There are lots of other possibilities, but maybe comparing md5's of  
tarballs would be a safer bet.

Does CVS not have a change number, or a changeset or commit ID that  
could be compare with a previous one?

________________________________________
Craig A. Berry
mailto:craigberry AT mac.com

"... getting out of a sonnet is much more
 difficult than getting in."
                 Brad Leithauser

Attachment: dist.patch
Description: Binary data