Re: Regression testing of tetex

[Junichi Uekawa <dancer@netfort.gr.jp>]

Hi,

> Did you notice that this sounds as if you were accusing Norbert of being
> lazy, and/or of bad package quality?  I don't think this is acceptable,
> without giving detailed reasons.  

I didn't meant to be insulting; it was some byproduct of my careless
choice of words. Apologies for any offence. I didn't mean it.

> - the tests discussed should be added upstream, or within the general
>   testing project I suggested - at least this is how I understood
>   Norbert

Yes, I agree that's the case. So there are two aspects to this
'automatic testing'

1.  Upstream point of view: extensively testing every style file etc,
    so that we have enough coverage and high-quality tex distribution.

2.  Debian point of view: assume upstream has already done some
    QA. Debian is adding some packaging scripts, and they need to be
    tested on top. To make sure the Debian-added portions work,
    implement some kind of 'package install -- build tex file -- check
    result' script.

Debian doesn't need an extensive scripting; we need some kind of
minimal test that checks for some fonts, some stylesheets, some
hyphenation, some dvi->ps->pdf dvi->pdf translation, so that what
we're generating under /etc/texmf and /var/lib/texmf are somewhat
correct.

> - Currently, only a few packages run automated tests upon building; even
>   less have added these in the Debian packaging

Yup. Except my packages and the packages I sponsor, which tend to have
testsuites. :P

> - AFAIK there's no testing framework yet, so that "building up tests
>   slowly" is more than just editing a few files, it means slowly typing
>   more and more commands (and waiting for them to finish, of course),
>   *and* inspecting the typeset output.

That happens to be the case even with some kind of testing framework.

regards,
	junichi
-- 
dancer@{debian.org,netfort.gr.jp}   Debian Project