Re: gnulib-v0.1: a non-event

[Jim Meyering]

On Mon, Oct 28, 2013 at 4:12 PM, Gary V. Vaughan <address@hidden> wrote:
> Hi Jim,
>
> On Oct 29, 2013, at 12:01 PM, Jim Meyering <address@hidden> wrote:
>> I have just pushed a signed v0.1 tag.  It holds no particular meaning.
>>
>> I find that smaller commit-count numbers are more reader-friendly
>> than the 8000+ numbers we had reached relative to the v0.0 patch.
>>
>> Thus, for the next 999 commits, git describe will print something like this:
>>
>>  v0.1-NNN-gHHHH...
>>
>> I.e., with no more than 3 digits in the commit count part.
>
> If we can find a way to automate the per-1000 commits tagging (presumably
> with a commit hook), then that lends itself to an easy solution to the
> shallow clone issue with bootstrap, by making sure the shallow clones are
> at least 1000 changesets deep.

The default is to require a GPG-signed tag (preferably using a
well-connected key), so it will not be done via a commit hook.
However, I could run a cron job that would remind me when we
get within say 50 commits of the limit, and then create/push the
signed tag manually.

However, isn't this based on the premise that shallow clones are
somehow useful?  Did you try the recommended procedure of
using a reference gnulib repository?  E.g.,

  ./bootstrap --gnulib-srcdir=../gnulib

The shallow-clone hack is to accommodate those who want to do
a quick one-off bootstrap; imho, not appropriate for those of us who
run it frequently, or as a maintainer.