[PATCH 3 of 3] hook: report untrusted hooks as failure (issue5110) (BC)

[timeless]

Indeed, much better.

On Sat, Apr 16, 2016 at 1:16 PM, Matt Mackall <mpm at selenic.com> wrote:
> On Fri, 2016-04-15 at 15:26 -0400, timeless wrote:
>> Pierre-Yves David wrote:
>> >
>> > timeless wrote:
>> > >
>> > > can i convince you to replace "denied" with "skipped"?
>> >
>> > Skipped seemed a bit too light for the abort case:
>> >
>> > Abort: pretxnopen.foo hook skipped (from unstrusted source)
>> >
>> > (it is not really "skipped" as it aborted the whole operation).
>> >
>> > What do you think?
>> I'm thinking "Can I get it in a sentence?"
>> But in practical terms, can you add tests w/ this commit showing
>> pretxnopen/pretxnclose (passing, failing)?
>>
>> "forbidden" is probably the right word in the pretxnopen case (instead
>> of denied).
>
> No, "forbidden" is not right either. This is instead something we're doing out
> of an abundance of caution for the user's benefit. I've tweaked the landed
> output thusly:
>
> --- /home/mpm/hg/tests/test-hook.t
> +++ /home/mpm/hg/tests/test-hook.t.err
> @@ -845,7 +845,7 @@
>    > txnclose.testing=echo txnclose hook called
>    > EOF
>    $ touch a && hg commit -Aqm a
> -  warning: txnclose hook forbidden (from untrusted config)
> +  warning: untrusted hook txnclose not executed
>    $ hg log
>    changeset:   0:3903775176ed
>    tag:         tip
> @@ -864,7 +864,8 @@
>    $ touch b && hg commit -Aqm a
>    transaction abort!
>    rollback completed
> -  abort: pretxnclose hook forbidden (from untrusted config)
> +  abort: untrusted hook pretxnclose not executed
> +  (see 'hg help config.trusted')
>    [255]
>    $ hg log
>    changeset:   0:3903775176ed
>
> Note the change of the pseudo-hint to a real abort hint.
>
> --
> Mathematics is the supreme nostalgia of our time.
>