[smc-discuss] PROPOSAL: A new developer workflow

Rajeesh K V rajeeshknambiar at gmail.com
Mon Aug 10 01:28:13 PDT 2015


On Mon, Aug 10, 2015 at 3:44 AM, Ashik Salahudeen <aashiks at aashiks.in> wrote:
> Hi all,
>
> We (Akshay and me) maintain SMC's font package for Arch Linux. We were
> looking at automating package updates and generally automating steps we take
> for new releases. A look at our current gitlab repos
> (https://gitlab.com/groups/smc) reveals that our workflows may not be
> entirely suitable for automation, including tests.

Nice!

>
> We explored setting up a project on a hosted Jenkins to build our fonts, and
> it turns out to be easy. We were thinking that if our development process
> (vbased on git) can be streamlined a bit, we can make things like automated
> testing easier, and will know when our tests fail and so on. We can move
> faster and better in this fashion. The proposal is not big, its just
> altering our git workflow a wee bit. The rest is setting up infraswtructure
> for automated build and release. As a starting point, I have put the key
> points in the document available at the link below.
>
> https://etherpad.mozilla.org/qIr3XSaieY
>
> Please go through and respond. If we have a consensus, we can setup a build
> server and add our projects in it.

I speak only for fonts development (not libindic, keyboard etc as they
have different requirements and use cases).
1.  Jenkins is a developer tool, with the purpose of catching build
and test failures.
2.  Font ≠ code. We cannot capture 'build failures' as technically ttf
generation by fontforge does not 'fail'.
3.  Testing is required to catch font failures. Rejah proposed test
automation, it would be great to incorporate automated
shaping/rendering testing incorporated.
4.  Our font repo received only a handful of commits now, so build per
commit - no nightly required
5.  No build from jenkins end up in user downloadable area, it would
be a manual activity to ensure quality and release readiness. When we
release fonts, people expect it to be 'final' - no incremental
improvement model is applicable here.

>
> Regards,
> Ashik Salahudeen
>


-- 
Rajeesh
http://rajeeshknambiar.wordpress.com



More information about the discuss mailing list