Optimisations and undefined behaviour

[Manuel López-Ibáñez]

On 10/11/15 11:56, David Brown wrote:
> That's why it's good that there are lots of people thinking together on
> these specifications, and not just one person!  Or perhaps we should all
> switch to using ARM chips...

The people who think and know a lot about these things are the members of the 
C/C++ standards committee. After this long discussion, I haven't seen anything 
new that is not mentioned already in the FAQ and links therein: 
https://gcc.gnu.org/wiki/FAQ#undefinedbut

and specially the last sentence is still true as ever:

"Perhaps we could, but not under the current Standard. [...] But the C Standard 
is under revision: perhaps, if this is important enough to you, you can 
convince the committee [...]. Good luck."

Of course, GCC should warn as much as possible about UB and behave as 
predictably and user-friendly as possible. Yet, many things in GCC work the way 
they work because nobody had the right approach, time or perseverance to fix 
them. Some of them may be too hard to "fix" or just create more problems. The 
way to figure that out is not by "debating" in a mailing list. It is by 
stretching your fingers and trying to fix them 
(https://gcc.gnu.org/wiki/GettingStarted#Basics:_Contributing_to_GCC_in_10_easy_steps). 


Unfortunately, I have learned the hard way myself that until you try to 
actually fix (fix as in implement something that can be added to mainline GCC 
and it is not a toy research project or pure mailing list speculation) those 
issues, you cannot really appreciate what needs to be done to fix them. Case in 
point: https://gcc.gnu.org/PR18501

Cheers,

Manuel.