Fix compiler warnings about uninitialised variables and issue #15

[robertgj]

1. Fix compiler warnings about uninitialised variables.

2. Fix issue results are not apparently deterministic? #15 "results are not apparently deterministic?" in blkchol2.c:
   --- sedumi/blkchol2.c 2020-08-15 23:10:22.858943847 +1000
   +++ SeDuMi_1_3/blkchol2.c 2020-08-15 23:09:49.000000000 +1000
   @@ -113,7 +113,7 @@
   ------------------------------------------------------- /
   xkk = x[inz];
   if(xkk > lb[k]){ / now xkk > 0 */

•  if(xkk < ub){
  

•  if((m>1) && (xkk < ub)){
     ubk = maxabs(x+inz+1,m-1) / maxu;
     if(xkk < ubk){
  

3. Apply fix to sedumi.m recommended in SparsePOP readme.txt (the second is already in):

NOTE4: SeDuMi in Github (2014-08-14) has two bugs to use it. We list
two modifications of SeDuMi:
(1) Remove % in the head of the 792nd line in sedumi.m
(2) Replace the 77th line in eigK.m by
lab(li+1:li+nl) = x(xi+1:xi+nl);

[siko1056]

I don't like this change. Please set j in a separate line before for clarity.

[siko1056]

Obviously not supported by clang (macOS) https://travis-ci.org/github/sqlp/sedumi/jobs/718183038

Is it too difficult to remove the unused variables as well? Then those flags are not needed.

[siko1056]

Thank you for your effort @robertgj . I reviewed your changes and suggest two changes.

[robertgj]

Respond to siko1056 requested changes

[siko1056]

Thank you again @robertgj for your changes. The reason for the macOS buildbot to fail is OpenMathLib/OpenBLAS#2783 and is no blocker for the merge.

Just one question before I merge: Where is the macro MEX_DEBUG documented for Matlab (I know it is given in Octave)? I think your intention is to avoid the unused variable warning in case mxAssert() statements are not compiled, right?

[robertgj]

I'm afraid I don't have access to MATLAB. Add -DMEX_DEBUG in the mex template to enable mxAssert checks when compiling MATLAB with install_sedumi.m.

[siko1056]

Having a closer look at it, I am convinced introducing the definition -DMEX_DEBUG is the wrong way to go. -Wall -Wextra are removed from install_sedumi() anyway, thus why complicating things now?

Can you strip your second commit (3d0e453) and just care about my two assertions for the first commit (313c7d2)?

1. Fix compiler warnings about uninitialised variables and results are not apparently deterministic? #15 (313c7d2)
2. Fix unused variable warnings when compiling SeDuMi (3d0e453)

Sorry, I don't want to delay things unnecessary, but quick fixes now are painful to deal with later.

[robertgj]

Revert previous change to robertgj/sedumi as requested by siko1056.

I am not sure what is happening with examples/README.md and examples/test_sedumi.m. They are the same in my clones of robertgj/sedumi and sqlp/sedumi.

1. blkchol2.c : do not call BLAS FORTRAN IDAMAX with an array of length 1
2. sedumi.m : add 'Octave:singular-matrix' warning id
3. maxstep.m, trydif.m, widelen.m : suppress octave-6 warning about implicit all
4. dpr1fact.c, bwblkslv.c, spscale.c fwblkslv.c urotorder.c,: suppress compiler
   warnings about unused and uninitialised variables
5. psdframeit.c, sqrtinv.c, symfctmex.c, vecsym.c: suppress compiler warnings
   about unused variables

[siko1056]

All changes look good, I care about the files I changed in the "examples" directory, if things go wrong.