Le Fri, 19 Jun 2015, Gabriel Scherer a écrit :
> This change to the placement of attributes that was discussed in
>   http://caml.inria.fr/mantis/view.php?id=6612
>   https://github.com/ocaml/ocaml/pull/152
> 
> It is indeed a mistake that it is not in the Changelog. As a meta-comment
> (not discussing this particular patch), in the future we might need to have
> "a proper Changelog line" as an explicit requirement for all submitted
> patches (I would be tempted to add "a testsuite test exercising the
> affected features"), because it is fairly error-prone to try to track
> changes to add to the changelog after they've been merged.
> 
> The rationale for this change is that the syntax was perceived as an
> improvement by the ppx-users that expressed an opinion on the issue, and
> they asked for this to be released as soon as possible hoping to facilitate
> the transition to the new syntax (few existing users).
> The change was tested against all released OPAM packages, and riak_ppx was
> the only package that broke. It seems that either (1) the code using the
> old syntax had not been released as OPAM packages yet (in retrospect it
> seems reasonable that using OPAM testing as only testing vehicle is rather
> fragile for very recent features; on the other hand, I don't know which
> other tests could be put in place) or (2) said packages did not break in a
> way that was perceptible to the automated tool.

ppx_deriving compiles with 4.02.2, but its tests do not pass. I indeed
have some unreleased code that breaks (it's on a distinct opam repo).
I suspect the tool does not run tests?

The syntax change might be an improvement indeed, but disabling the old behavior
(attaching attributes to record labels) is *breaking* in a way that
makes fixes difficult. For instance, introducing 'nonrec' is breaking,
but in an easy to fix way; the attribute change is hard (you need to
edit code manually to move annotations).

> 
> I think 4.02.2 is a slightly less minor release than is usual: it does have
> some new features that are not simply bugfixes, because they were requested
> by users which were eager to see them released (typically all the
> ppx-related changes). This is a delicate compromise to make, and
> unfortunately there are indeed a couple regressions -- despite a very
> silent rc1-to-release period. The only other one I know of is
> http://caml.inria.fr/mantis/view.php?id=6908 .


-- 
Simon

http://weusepgp.info/
key 49AA62B6, fingerprint 949F EB87 8F06 59C6 D7D3  7D8D 4AC0 1D08 49AA 62B6