Enabling some compiler warnings produce many that could be fixed

[Jaaaky]

I'm trying to port your library, enabling compiler warnings produces many that could be fixed and actually I think fixed them would enhance the quality of the lib too much.

Here are most of them

../../libraries/nlopt/plis.c: In function 'plis_':
../../libraries/nlopt/plis.c:189:5: error: floating constant exceeds range of 'float' [-Werror=overflow]
  189 |     eta9 = 1e120;
      |     ^~~~
compilation terminated due to -Wfatal-errors.
cc1: some warnings being treated as errors



../../libraries/nlopt/newuoa.c: In function 'trsapp_':
../../libraries/nlopt/newuoa.c:245:12: warning: comparing floating-point with '==' or '!=' is unsafe [-Wfloat-equal]
  245 |     if (dd == zero) {
      |            ^~
../../libraries/nlopt/newuoa.c:406:24: warning: comparing floating-point with '==' or '!=' is unsafe [-Wfloat-equal]
  406 |     if ((double) isave == zero) {
      |                        ^~
../../libraries/nlopt/newuoa.c:413:15: warning: comparing floating-point with '==' or '!=' is unsafe [-Wfloat-equal]
  413 |     if (tempa != tempb) {
      |               ^~
../../libraries/nlopt/newuoa.c: In function 'bigden_':
../../libraries/nlopt/newuoa.c:643:14: warning: comparing floating-point with '==' or '!=' is unsafe [-Wfloat-equal]
  643 |   if (sstemp == 0) return NLOPT_ROUNDOFF_LIMITED;
      |              ^~
../../libraries/nlopt/newuoa.c:897:15: warning: comparing floating-point with '==' or '!=' is unsafe [-Wfloat-equal]
  897 |     if (tempa != tempb) {
      |               ^~
../../libraries/nlopt/newuoa.c: In function 'biglag_':
../../libraries/nlopt/newuoa.c:1362:15: warning: comparing floating-point with '==' or '!=' is unsafe [-Wfloat-equal]
 1362 |     if (tempa != tempb) {
      |               ^~
../../libraries/nlopt/newuoa.c: In function 'update_':
../../libraries/nlopt/newuoa.c:1443:41: warning: comparing floating-point with '==' or '!=' is unsafe [-Wfloat-equal]
 1443 |  } else if (zmat[*knew + j * zmat_dim1] != zero) {
      |                                         ^~
../../libraries/nlopt/newuoa.c: In function 'newuob_':
../../libraries/nlopt/newuoa.c:2090:11: warning: comparing floating-point with '==' or '!=' is unsafe [-Wfloat-equal]
 2090 |  if (d__1 == 0) { rc = NLOPT_ROUNDOFF_LIMITED; goto L530; }
      |           ^~
../../libraries/nlopt/newuoa.c:2325:29: warning: comparing floating-point with '==' or '!=' is unsafe [-Wfloat-equal]
 2325 |     if (ksave == 0 && delta == rho) {
      |                             ^~

[Jaaaky]

Here are some compiler flags to enable to show dozen of possible problems:

 -Wall -Wextra  -Werror=overflow -Wtype-limits -Woverflow -Wunused-value -Wmaybe-uninitialized -Werror=type-limits -Werror=maybe-uninitialized -Wfloat-equal -Wmissing-declarations -Werror=deprecated-declarations -Wincompatible-pointer-types -Wint-conversion

Solving those can help avoid dozen of bugs and undefined behaviors

[Jaaaky]

Some more warning that could be fixed, but needs logic checking that I could not do

ccsa_quadratic.c: In function 'ccsa_quadratic_minimize':
ccsa_quadratic.c:365:4: warning: floating constant exceeds range of 'float' [-Woverflow]
  365 |    for (i = 0; i < m; ++i) dual_ub[i] = 1e40;
      |    ^~~

plip.c: In function 'plip_':
plip.c:204:5: warning: floating constant exceeds range of 'float' [-Woverflow]
  204 |     eta9 = 1e120;
      |     ^~~~

pnet.c: In function 'pnet_':
pnet.c:224:5: warning: floating constant exceeds range of 'float' [-Woverflow]
  224 |     eta9 = 1e120;
      |     ^~~~

mma.c: In function 'mma_minimize':
mma.c:241:4: warning: floating constant exceeds range of 'float' [-Woverflow]
  241 |    for (i = 0; i < m; ++i) dual_ub[i] = 1e40;
      |    ^~~

DIRsubrout.c: In function 'direct_dirheader_':
DIRsubrout.c:1490:2: warning: floating constant exceeds range of 'float' [-Woverflow]
 1490 |  *epsfix = 1e100;
      |  ^
DIRsubrout.c: In function 'direct_dirsummary_':
DIRsubrout.c:1586:3: warning: floating constant exceeds range of 'float' [-Woverflow]
 1586 |   if (*fglobal > -1e99)
      |   ^~

plis.c: In function 'plis_':
plis.c:189:5: warning: floating constant exceeds range of 'float' [-Woverflow]
  189 |     eta9 = 1e120;
      |     ^~~~

[Jaaaky]

This looks like serious one too:

nldrmd.c:205:11: warning: ‘memset’ specified bound between 18446744056529682432 and 18446744073709551608 exceeds maximum object size 9223372036854775807 [-Wstringop-overflow=]
  205 |           memset(xcur, 0, sizeof(double)*n);
      |           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

[Jaaaky]

Another interesting one

bobyqa.c:1037:19: error: 'adelt' may be used uninitialized in this function [-Werror=maybe-uninitialized]
 1037 |     temp = *adelt * *adelt - wfixsq;
      |            ~~~~~~~^~~~~~~~
compilation terminated due to -Wfatal-errors.
cc1: some warnings being treated as errors

[stevengj]

Are you compiling on a platform where float is the same as double? The floating constant exceeds range of 'float' make no sense otherwise, since the indicated constants are within the range of double precision.

[stevengj]

warning: ‘memset’ specified bound between 18446744056529682432 and 18446744073709551608 exceeds maximum object size 9223372036854775807 [-Wstringop-overflow=]
  205 |           memset(xcur, 0, sizeof(double)*n);

The code looks correct to me. I have no idea what this warning means — if n is the size of the array, as it is in this code, then this is setting the correct number of bytes to zero.