fix BLAS usage (broken by #53)

[devernay]

As pointed out by @jschueller in #66, PR #53 broke the BLAS support by not passing all arguments as references.

[devernay]

EDIT: This is resolved

Compiles, but tests don't pass, see https://github.com/devernay/cminpack/actions/runs/9693835725/job/26750173978?pr=68

failure example:

--- examples/ref/tlmder1c.ref   2024-05-28 22:47:05
+++ examples/tlmder1c.out   2024-06-27 11:06:38
@@ -1,7 +1,7 @@
-      final l2 norm of the residuals     0.09063596
+      final l2 norm of the residuals             15

-      exit parameter                         1
+      exit parameter                         2

       final approximate solution

-          0.08241058       1.133037       2.343695
+                   1              1              1

[jschueller]

tests pass on my end

[anntzer]

I no longer regularly use cminpack and am furthermore unable to review this in the coming month or so due to other priorities, so feel free to push this forward without me. I'll keep a note to look at this if this is still open once I'm available again, though.

[devernay]

tlmdifc failed because of FP precision. disabled it

[jschueller]

do you plan a new patch release ?

[devernay]

Once we settle wether https://github.com/devernay/cminpack/pull/64 can be merged. Can you review it please?

[jschueller]

yes, its not so great atm see #71

[jschueller]

@devernay would it be possible to release a new version so that the fixes for using blas are included ? it would avoid downstream to use the broken 1.3.9 version when blas is enabled like homebrew does

[devernay]

@devernay would it be possible to release a new version so that the fixes for using blas are included ? it would avoid downstream to use the broken 1.3.9 version when blas is enabled like homebrew does

done!