Re: [commit 1/3] Import gnulib's update-copyright script

[Pedro Alves <palves at redhat dot com>]

On 04/18/2012 03:52 PM, Joel Brobecker wrote:

>> I don't see how that scripts would have prevented the selective
>> check-in.  :-)
> 
> Perhaps not, but what I would have done was doing an update first,
> check that in, and then add the new module. WDYT?


Yes, that or the other way around, is the right way to do update gnulib.
We should never end up with a mix of files from different versions,
as that gets messy and undone with the next import.  See how Yao imported
inttypes recently -- he imported it from that same 2010 version.

> 
>> It may help by keeping the git version in some script variable, that
>> the script checks, instead of having to fetch the version from the
>> ChangeLog?  The new gnulib/ parent directory seems like a good place
>> for this stuff.
> 
> Sure. Since there does not appear to be gnulib releases, I was
> under the impression that we'd always go with the current head,
> somehow. 


Yeah, but of the whole gnulib as a unit.  There's a lot of dependencies
between the modules.  And there's no telling if some random import
would break some host, due to some gnulib, so it's best if we know
exactly which version was used.

> Having the hash in the script is a good idea, makes
> things completely reproduceable...


Exactly.

Okay, we're in violent agreement at this point.  :-)

I'll go update gnulib to the current version.

-- 
Pedro Alves