On Sun, Aug 12, 2018 at 6:43 PM, Sara Golemon <poll...@php.net> wrote:
> On Sun, Aug 12, 2018 at 12:47 AM, Stanislav Malyshev > <smalys...@gmail.com> wrote: > > Undefined behavior is undefined :) > > > (Ignoring your followup for a moment) > Even for undefined behavior, we should *try* to make the behavior > repeatable. I know we wouldn't need to, but we should always go for > least surprise. > It's the side-effect hiding in the array cast from print_r/var_dump or > certain other presumable "read-only" actions causing the > interpretation of of the object compare that's maximum surprise. > > >> <https://github.com/php/php-langspec/blob/master/spec/10- > expressions.md#user-content-relational-operators> > > > > So looks like this part of the spec is violated, and I stand corrected - > > it's actually defined in this case, and should be fixed to follow the > > spec. I still think it makes little sense, but we have the spec, so we > > have to follow the spec. > > > > In this case, it's ok to fix it in all active versions - the spec wins I > > think. Or, we have to change the spec :) > > > Personally, I vote for fix on active branches, but it's not an urgent > bug, so I'll give some more time for feedback before pushing the PR > below. > I'd recommend against fixing active branches -- this is quite likely to break some tests using sort() somewhere (even if there is no resulting functional change), while the fix itself is probably very niche. Nikita > > Depends on what the "fix" is, I assume. > > > I went with https://github.com/php/php-src/pull/3434 which has the > same overall complexity, though several more operations per element so > it'll be nominally less performant. Alternate implementations > welcome. > > -Sara > > -- > PHP Internals - PHP Runtime Development Mailing List > To unsubscribe, visit: http://www.php.net/unsub.php > >
