13-06-2016, 04:56 AM
According to Rule 4-5-1 sizeof(bool_type_variable) is not allowed, although it is very meaningful in serialization.
Is this intentional?
If so then why?
Is this intentional?
If so then why?
13-06-2016, 04:56 AM
13-06-2016, 09:31 AM
I think this is because sizeof(bool) is implementation-defined according to ISO C++03 chapter 5.3.3.
If you want to code a portable serialization function, you have to store it in an integer with a guaranteed fixed size across all platforms (e.g. using uint8_t from stdint.h or your own type header)
If you want to code a portable serialization function, you have to store it in an integer with a guaranteed fixed size across all platforms (e.g. using uint8_t from stdint.h or your own type header)
13-06-2016, 10:05 AM
Thanks for the reply, but:
1. Still the use of cstdint types is only advisory. The same problem exists for all built-in types, so why bool should be different? The use of typedefs for bool, could be advisory as well (if that's the intention).
2. The rationale in this section, referes only to meaningless and therfore error-prone. The use of sizeof is not meaningless.
3. The portability of types can be verified with static_assert (or otherwise equivalent implementations of it).
By not addressing this issue directly in the standard, it becomes easier for me to ignore it as an advisory or overlooked.
As I do take meaures to avoid copmiler differences, where it might affect our S/W, my problem is that the rule for this in the static analysis tool we use, makes it harder for me to spot what I consider the "real" violations of this rule.
1. Still the use of cstdint types is only advisory. The same problem exists for all built-in types, so why bool should be different? The use of typedefs for bool, could be advisory as well (if that's the intention).
2. The rationale in this section, referes only to meaningless and therfore error-prone. The use of sizeof is not meaningless.
3. The portability of types can be verified with static_assert (or otherwise equivalent implementations of it).
By not addressing this issue directly in the standard, it becomes easier for me to ignore it as an advisory or overlooked.
As I do take meaures to avoid copmiler differences, where it might affect our S/W, my problem is that the rule for this in the static analysis tool we use, makes it harder for me to spot what I consider the "real" violations of this rule.
31-10-2016, 03:12 PM
You're right that the C++ standard does describe sizeof as an operator, and sizeof(bool) is a well defined operation so ought to be allowed
This will be addressed in a future edition of the rules
This will be addressed in a future edition of the rules