Repository navigation
CWG3142 [cpp.predefined] Binary integer literal and digit separators for __LINE__ #826
Description
Activity
- changed the title
[-][cpp.predefined] Binary integer literal and digit separators for `__LINE__`[/-][+]CWG3142 [cpp.predefined] Binary integer literal and digit separators for `__LINE__`[/+]on Dec 6, 2025 Perhaps SG22 and WG14 should also see this, as
- wording for
__LINE__(not restricting form of integer literal/constant) was copied from C, and - digit separators and binary integer literals have been copied to C23 (via WG14 N2549 and WG14 N2626).
- wording for
Feel free to send this to WG14.
It's possible to
static_assertthat__LINE__expands to a literal of expected forms since C++11 (example)There is an even easier way to assert this (for the form that all compilers use):
#line __LINE__This requires a decimal integer literal, possibly with digit separators. Note that it will even interpret
010as ten instead of eight. A footnote in the C standard even mentions this without saying anything about it being possibly invalid, so presumably it's intended to be valid.Additionally, digit separators being present would break the common use case of concatenating __LINE__ with an identifier:
#define CAT(X,Y)CAT_(X,Y) #define CAT_(X,Y)X##Y #define FOO()int CAT(x,__LINE__) FOO(); FOO(); //each declaration gets a unique name from __LINE__ //but something like x1'2 isn't a valid identifier
See also C issue 1016 which discusses this.
See also C issue 1016 which discusses this.
Yeah. I found this several days ago and attempted to mail WG14 reflector to expand the discussion for this issue.
C issue 1016 was recently resolved but the resolution didn't touch binary integer literals.
C issue 1016 was recently resolved but the resolution didn't touch binary integer literals.
I have created Issue 1029 which should resolve the other issues.
Full name of submitter (unless configured in github; will be published with the issue): Jiang An
Reference (section label): [cpp.predefined]
Link to reflector thread (if any):
Issue description:
Discovered in cplusplus/draft#8584 (comment).
It's possible to
static_assertthat__LINE__expands to a literal of expected forms since C++11 (example), or assert this at run time even in ancient C++98. And then in C++14, we introduced binary integer literals and digit separators, which allows__LINE__to expand to a literal of a formerly unexpected form.As a result, if an implementation switch to use the new forms of integer literals for
__LINE__, some valid C++11 program may become ill-formed or have behavior changed.Should there be an Annex C entry for this, or should we restrict the forms for
__LINE__?Suggested resolution: