[Rivet] Move bootstrap into main repository for versioning?

[Frank Siegert]

Actually, these two issues just got me thinking whether it would be
wise to move the bootstrap script into Rivet main such that it can
remain working for older versions. The download would then be from
something like

http://rivet.hepforge.org/hg/rivet/raw-file/rivet-2.0.0/rivet-2-bootstrap

and would simply remain as it is even if we have to change something for 2.1.0.

What do you think?

Cheers,
Frank

On 6 February 2014 10:00, Frank Siegert <frank.siegert at cern.ch> wrote:
> Hi Andy,
>
>>>> Frank, I've merged some extra changes into the bootstrap scripts, to move the autotools building into the dev mode only. When we release 2.0.1 do you want to remove mention of yaml-cpp completely, or just set INSTALL_YAMLCPP=0 by default?
>>>
>>> I would do the latter, and remove it at some later point, just to make
>>> it possible for users to install 2.0.0 at all.
>>
>> Ok. So it will require some special treatment to apply --with-yaml-cpp
>> if version < 2.1.0, otherwise without :-/
>
> Ah, right, I forgot about that. I would suggest to remove the
> --with-yaml_cpp from the configure and set INSTALL_YAMLCPP=0, but
> leave the installation block in, such that it's a simple one-line
> change in case somebody wants to install 2.0.0. When we release 2.2.0
> I think we can remove it.
>
>>> One more point about the "getting started" for new users: would you
>>> mind replacing the URL of the "Get Rivet" link on the homepage such
>>> that it points to the GettingStarted wiki page? I think currently it
>>> might be a bit misleading for users to be transferred directly to the
>>> download page.
>>
>> Done! Thanks.
>
> Thanks!
> (I hope users will notice the distinction between "Get Started" and
> "Download source code".)
>
>> However, the bootstrap scripts need to install cmake if it's not
>> found: it's needed to build Rivet due to the built-in yaml-cpp, and the
>> build from rivet-2-bootstrap-lcg currently fails on lxplus because there
>> isn't a system cmake installation (really?!?)
>> [I guess hg could be reverted to install on failed autodetection rather
>> than explicit request, since I don't think there is any version
>> sensitivity -- unlike for Cython.]
>
> Right, I'll update the bootstrap script now such that there is cmake
> autodiscovery/install also for the release version.
>
> Cheers,
> Frank