Re: [Tails-dev] Releasing automated tests

Delete this message

Reply to this message
Autore: bertagaz
Data:  
To: The Tails public development discussion list
Oggetto: Re: [Tails-dev] Releasing automated tests
Hi,

Sorry for the late reply.

On Wed, Oct 28, 2015 at 04:52:31PM +0000, anonym wrote:
> sajolida:
> > anonym:
> >> sajolida:
> >>>>> From: intrigeri <intrigeri@???>
> >>>>> Date: Tue, 20 Oct 2015 13:08:31 +0200
> >>>>>
> >>>>> FTR I dislike the idea of blacklisting such branches based on their
> >>>>> name. I'm not going to debate it here [...]


Intrigeri, what's on your opinion on the proposed way to blacklist them
made in this thread by anonym and sajolida (see below)? It's not based
on their name anymore.

> >>> I also don't think they should be tested. Maybe you could exclude them
> >>> from testing by their diff to their base branch. If all the diff is
> >>> under wiki/src then don't test that branch.


That seems to be the consensus here then, unless someone else raises.

> Then I think we can combine the "..." operator with another fancy Git
> feature I recently found, namely Git pathspec "magic signatures". So we
> could do:
>
>    BASE_BRANCH_DIFF="$(git diff $base_branch...$commit  -- \
>                            '*' \
>                            ':!/wiki' \
>                            ':!/ikiwiki.setup' \
>                            ':!/ikiwiki-cgi.setup')"
>    if [ -z "${BASE_BRANCH_DIFF}" ]; then
>        CUCUMBER_ARGS="${CUCUMBER_ARGS} --tag @doc"
>    fi

>
> where $commit is the commit we test, before merging the base branch
> locally. Interesting!


Agree, I like this way to encode it. Sounds at least like a nice baby
step to begin with, simple and effective.

This whole thread remembered me the discussion we had earlier about
feature branches, and made me think (again) we could as well use the
cucumber tag facility to lighten the number of tests: if building from a
feature|test branch, run with the --tag $branch_name, so developers can
choose which test they want their branch to be tested against. May be
useful for some branches were we know that it'll break some tests. But
we dropped this idea.

So if everyone agree on the proposal made here by anonym and sajolida, I
guess that's now a new "Code" ticket.

bert.