Re: [tap] RFC Status?

[Leon Timmermans <[e-mail]>]

On Wed, Sep 17, 2014 at 9:08 PM, Ovid <[e-mail]> wrote:

> A few notes on this.
>
> For subtests, they are indented four spaces. Always four spaces. If they
> are indented eight spaces, you know they're a subtest of a subtest
> (assuming there's an intervening subtest. That means this is valid TAP, but
> there is no subtest because it's indented too far:
>
> That is very useful to know, makes life for the consumer much easier.

> ok 1 - foo
>         ok 1 - bar indented too far
> ok 2 - bar passed
>
> The "ok 1 - bar" line would be discarded as "unknown".
>

I think I agree, though one could argue that's an unknown subline of a
subtest.


> So if, in a stream of TAP, you find something which would be valid tap if
> it were unindented by four spaces, it's a subtest. The trailing,
> unindented summary line must mirror the status of the subtest. Thus, the
> following is invalid:
>
> ok 1 - foo
>     not ok 1 - bar indented too far
> ok 2 - bar passed
>
> Currently, Test::Harness <http://search.cpan.org/dist/Test-Harness/>
> treats the above as passing when it should not. I believe Schwern filed a
> bug for that. I know Andy Armstrong and I both took a quick swing at
> extending Test::Harness but discovered that it was more difficult than it
> appeared.
>

I'm currently giving a parse error in that condition :-)

As for the YAMLish structured diagnostics, those were never official, as
> far as I recall. There were several disputes, one of which was whether or
> not we should use YAML or JSON (I recall even XML being mentioned). Some
> argued that any structured diagnostic should be allowed, but I reject that
> firmly because that makes life hell for the parser. If you asked me what my
> stance is today, I would think that structured diagnostics should valid
> JSON, but that raises questions about whitespace issues, UTF-8, and so on.
> If we could agree on the format, then we'd need to agree on the keys in the
> structured diagnostic dictionary (and how to extend them).
>

Deciding on the keys is bikeshedding, deciding on the format decidedly
isn't. To be honest I'm not even sure what would be good or bad at this
stage, though I can agree with not accepting every format and their cousin.

Leon