https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115990
gnzlbg <gonzalo.gadeschi at gmail dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |gonzalo.gadeschi at gmail dot
com
--- Comment #11 from gnzlbg <gonzalo.gadeschi at gmail dot com> ---
> so people were used to this.
I agree that some people are used to -Ofast and are aware of what it does (I
think I am one of those!). Those people will be fine if this API gets better.
The users that motivate doing the change are new users, those who are starting
to program today, which are not aware of this one gotcha (learning about it is
a rite of passage for compiler users).
> People will never read the documentation either way anyways.
I agree that people will never read the documentation, which is why I think
that it would be nice if this API - given its consequences - could be made a
bit more resilient against that. Currently, its interface, -Ofast, doesn't hint
me as a user that it will in any way deviate from all other -O_ options and
change the semantics of my program in non-compliant ways. The fact that other
-O_ options don't do that, helps build a mental model that none of the -O_
options do that, which is incorrect.
So I think that a more explicit API that makes these gotchas clearer would be
an improvement, as long as other valuable features like compatibility between
GCC and LLVM is not sacrificed.
Not sure what such an API would look like.
