Compile failure in rtapi/rtapi_math/e_asin.c

[ArcEye]

Issue by mbuesch Thu Aug 11 21:10:49 2016 Originally opened as https://github.com/machinekit/machinekit/issues/1035

---

Compiling rtapi/rtapi_math/e_asin.c
rtapi/rtapi_math/e_asin.c: In function ‘__ieee754_asin’:
rtapi/rtapi_math/e_asin.c:98:8: error: this ‘else’ clause does not guard... [-Werror=misleading-indentation]
      } else
        ^~~~
rtapi/rtapi_math/e_asin.c:100:3: note: ...this statement, but the latter is misleadingly indented as if it is guarded by the ‘else’
   p = t*(pS0+t*(pS1+t*(pS2+t*(pS3+t*(pS4+t*pS5)))));
   ^
cc1: all warnings being treated as errors
Makefile:391: recipe for target 'objects/rtapi/rtapi_math/e_asin.o' failed

I'm not sure what the correct code would be. If the indent is correct (and thus the code is incorrect), asin seems to generate incorrect values for very small inputs.

[ArcEye]

Comment by mhaberler Tue Aug 16 10:07:55 2016

---

I think @strahlex imported this code

[ArcEye]

Comment by mhaberler Wed Aug 17 17:30:58 2016

---

@strahlex : I cannot even find any reference from where this code originated - could you help reverse-engineer/document the whereabouts?

the code fragment @mbuesch alluded to definitely looks fishy, and it would be valuable now to have some reference to the code base

[ArcEye]

Comment by mbuesch Wed Aug 17 17:36:13 2016

---

It seems to originate from an ancient BSD codebase http://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libm/src/e_asin.c

[ArcEye]

Comment by mhaberler Wed Aug 17 18:26:34 2016

---

there seem to be all sorts of variations of this file, and different attempts at patching this up:

https://android.googlesource.com/platform/bionic/+/7581f9c/libm/upstream-freebsd/lib/msun/src/e_asin.c http://www.scs.stanford.edu/histar/src/pkg/uclibc/libm/e_asin.c https://bugzilla.mozilla.org/show_bug.cgi?id=933257 05-fix-else-branch.patch https://github.com/freebsd/freebsd/commit/d95c118b83b9227e11b22b271e9a8fb9f29a0ed7 http://lists.uclibc.org/pipermail/uclibc/2002-May/024269.html https://bugzilla.mozilla.org/show_bug.cgi?id=405053

[ArcEye]

Comment by mbuesch Wed Aug 17 20:13:32 2016

---

Why is this needed anyway? Can't we replace it by something sane?

[ArcEye]

Comment by mhaberler Thu Aug 18 20:44:14 2016

---

absolutely - need to find a trustable source first ;)

[ArcEye]

Comment by machinekoder Fri Aug 19 18:06:50 2016

---

It's part of the libm source code and yes it is absolutely crazy code. If you have a better working version feel free to fix it. We introduced this file because the sin implementation of the libm shipped with Debian causes real-time delays for certain values. Ideal replacement would be something with deterministic processing time.

[ArcEye]

Comment by machinekoder Fri Aug 19 18:08:15 2016

---

See also https://github.com/machinekit/machinekit/issues/629