Poll on short circuiting

[Janus Weil]

Dear gfortran community,

since our recent discussions on short-circuiting did not really come
to a conclusive end (and we mostly just heard two loud minorities up
to now), I thought it would be nice to hear what the silent majority
thinks. Therefore I opened a quick doodle poll on this topic:

https://doodle.com/poll/gisivemyrpnkqnpm

I'd like to invite everyone who has an opinion on this to share it
with us, no matter if you're a contributor or a user of gfortran.
Plenty of arguments have been made for all the different options, see
the references below.

I'm not sure if this will be helpful at all, but I think it's worth a
try. We'll see what comes out ...

Cheers,
Janus



References:
[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57160
[2] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85599
[3] https://gcc.gnu.org/ml/fortran/2018-06/msg00065.html (plus follow-ups)
[4] https://gcc.gnu.org/ml/fortran/2018-07/msg00017.html (plus follow-ups)
[5] https://gcc.gnu.org/ml/fortran/2018-07/msg00086.html (plus follow-ups)
[6] some never-ending c.l.f. discussions

[Janus Weil]

Thanks to everyone who participated so far. At this point it seems
that the results reflect the controversy of the debate we had.

Note that I just edited the entries of those people who chose the
"always" option, checking all those boxes which are implied by
"always" (so that the poll is easier to read vertically). In hindsight
I should probably have left out the options "always" and "never",
which are representable by combinations of the other options.

Since there are no really clear majorities, I guess we don't really
learn a lot from this so far, except for some rather points:
* Everyone wants to do short-circuiting in some form (no one chose "never").
* The people who don't care didn't participate.
* Almost everyone wants to do short-circuiting with pure functions.

However, it seems many gfortran contributors have not participated
yet. I hope some people will still do it.

Cheers,
Janus

Post by Janus Weil
Dear gfortran community,
since our recent discussions on short-circuiting did not really come
to a conclusive end (and we mostly just heard two loud minorities up
to now), I thought it would be nice to hear what the silent majority
https://doodle.com/poll/gisivemyrpnkqnpm
I'd like to invite everyone who has an opinion on this to share it
with us, no matter if you're a contributor or a user of gfortran.
Plenty of arguments have been made for all the different options, see
the references below.
I'm not sure if this will be helpful at all, but I think it's worth a
try. We'll see what comes out ...
Cheers,
Janus
[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57160
[2] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85599
[3] https://gcc.gnu.org/ml/fortran/2018-06/msg00065.html (plus follow-ups)
[4] https://gcc.gnu.org/ml/fortran/2018-07/msg00017.html (plus follow-ups)
[5] https://gcc.gnu.org/ml/fortran/2018-07/msg00086.html (plus follow-ups)
[6] some never-ending c.l.f. discussions

[Janus Weil]

So, since there hasn't been too much activity on the poll in the last
days, a little intermediate summary:
* 18 people have participated in the poll so far.
* 7 people voted for doing short-circuiting only at -O1 and above.
* 7 people voted for always doing short-circuiting.
* 3 people voted for doing short-circuiting only with pure functions
(without distinction of compiler options)
* 1 person voted for short-circuiting at -O1, but nowhere else (which
doesn't make a whole lot of sense to me).

Regarding the central issue of PR 57160 (whether to do
short-circuiting at -O0), we basically have a tie right now.

Some notes:
* In case someone wants to change/refine his vote, that is still possible.
* It is also possible to use the "ifneedbe" option (yellow check mark)
to indicate that you are undecided/agnostic on some option.

I'm not sure if there will be any further participants, but I'll leave
the poll open just in case. Anyway, it seems we will not have any
clear majorities, so I guess I'll just stop wasting my time here.

Cheers.
Janus

Post by Janus Weil
Thanks to everyone who participated so far. At this point it seems
that the results reflect the controversy of the debate we had.
Note that I just edited the entries of those people who chose the
"always" option, checking all those boxes which are implied by
"always" (so that the poll is easier to read vertically). In hindsight
I should probably have left out the options "always" and "never",
which are representable by combinations of the other options.
Since there are no really clear majorities, I guess we don't really
* Everyone wants to do short-circuiting in some form (no one chose "never").
* The people who don't care didn't participate.
* Almost everyone wants to do short-circuiting with pure functions.
However, it seems many gfortran contributors have not participated
yet. I hope some people will still do it.
Cheers,
Janus

Post by Janus Weil
Dear gfortran community,
since our recent discussions on short-circuiting did not really come
to a conclusive end (and we mostly just heard two loud minorities up
to now), I thought it would be nice to hear what the silent majority
https://doodle.com/poll/gisivemyrpnkqnpm
I'd like to invite everyone who has an opinion on this to share it
with us, no matter if you're a contributor or a user of gfortran.
Plenty of arguments have been made for all the different options, see
the references below.
I'm not sure if this will be helpful at all, but I think it's worth a
try. We'll see what comes out ...
Cheers,
Janus
[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57160
[2] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85599
[3] https://gcc.gnu.org/ml/fortran/2018-06/msg00065.html (plus follow-ups)
[4] https://gcc.gnu.org/ml/fortran/2018-07/msg00017.html (plus follow-ups)
[5] https://gcc.gnu.org/ml/fortran/2018-07/msg00086.html (plus follow-ups)
[6] some never-ending c.l.f. discussions