Submitter: Fred Tydeman (USA)
Submission Date: 2004-08-13
Source: WG 14
Reference Document: Version: 1.0
Subject:Is cabs() a type-generic macro?
The standard is not clear as to which type-generic macro(s) should be used to compute the absolute value of real, complex, and imaginary types.
Here are two viewpoints.
1) cabs() is not a type-generic macro. Instead fabs() covers fabsf(), fabs(), fabsl(), cabsf(), cabs(), and cabsl().
If this viewpoint is correct, then 7.22p2 may need to be updated along the lines of "except modf and cabs". Also, in 7.22p7 (3rd line from end), cabs(fc) should be changed to fabs(fc).
2) cabs() is a type-generic macro (but only for complex arguments). That is cabs() covers cabsf(), cabs(), and cabsl(). In addition, fabs() covers fabsf(), fabs(), fabsl(), cabsf(), cabs(), and cabsl().
If this viewpoint is correct, then 7.22p6 needs to be updated with cabs. Also, B.21 needs to have cabs added to it.
It seems clear that cabs() is not a type-generic macro.
In addition, it is not clear what modf is in 7.22p5.
Suggested Technical Corrigendum
7.22p5 Add ", except modf" after "<math.h>"
7.22p7 (3rd line from end), cabs(fc) should be changed to fabs(fc).
7.22p2 says that every function in <math.h> and <complex.h> without an l or f suffix has a corresponding type-generic macro. Except modf.
However, 7.22p2 says nothing about what the macro is called.
7.22p4 says that if there is a function F() in <math.h> and a corresponding function cF() in <complex.h>, then the TG-macro for both F() and cF() is called "F". In addition, the TG-macro for both fabs() and cabs() is called "fabs".
7.22p5 and 7.22p6 say that for all remaining functions that have TG-macros, the macro has the same name as the function.
However, none of 7.22p4 to 7.22p6 say which functions have TG-macros.
Adding "except cabs" to 7.22p2 would directly contradict 7.22p4, because it says that there is no TG-macro corresponding to cabs (and, therefore, fabs).
If there's a defect in the wording, it's that 7.22p5 should say "except modf" after "<math.h>". Or all of 7.22p4 to 7.22p6 should have something like "Where a function has a corresponding type-generic macro" inserted in them. But we don't even believe that is necessary.
To get the absolute value of a float complex using a type-generic macro, use fabs. Therefore, 7.22p7 (3rd line from end) needs to change cabs(fc) to fabs(fc).
As G.7p1 correctly shows, to take the absolute value of an imaginary type, use fabs.
Previous Defect Report < - > Next Defect Report