Thanks for your explanation. I do understand now how '1 << 31' could cause a potential problem.
On Fri, Jan 8, 2021 at 6:10 AM Nico Huber <[email protected]> wrote: > Hi Harshit, > > On 08.01.21 01:35, Harshit Sharma wrote: > > Although, I agree that 1u << 31 is much better, I believe 1 << 31 is not > > wrong either as long as it is assigned to an 'unsigned int' as the > compiler > > performs an implicit conversion from a lower data type to a higher data > > type ('int' to 'unsigned int' in this case). > > you are right about the implicit conversion, but this happens too > late. The actual error may happen before the conversion. I guess > mostly because of the way signed numbers are handled in assembler, > compilers usually produce code that results in -2^31 for `1 << 31`. > This -2^31 then implicitly converted to an unsigned is 2^31 which > is what we usually intend. > > But the theoretical problem is what happens if the compiler produces > code that does not result in -2^31 for the signed number. As the C > standard implicitly says `1 << 31` is undefined, the compiler can > basically do what it wants and would still be C compliant. > > Nico >
_______________________________________________ coreboot mailing list -- [email protected] To unsubscribe send an email to [email protected]
