Control: tags -1 patch Control: tags -1 pending I think the memory leak results from a missing cleanup in the mips64el specific path. The full patch (including the original one) is here.
I am going to upload the new version ASAP. Best Ole
Author: Lars Kristian Lundin <[email protected]> Description: Fix testing on mips64el Bug: #753089 Url: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=753089#10 --- a/cplcore/tests/cpl_matrix-test.c +++ b/cplcore/tests/cpl_matrix-test.c @@ -200,6 +200,7 @@ FILE * stream; const double * data; cpl_boolean do_bench; + cpl_boolean did_fail; cpl_error_code code; const cpl_matrix * null; /* This one is expected to be NULL */ @@ -1579,6 +1580,8 @@ cpl_msg_info("", "Try to solve increasingly large systems A^TAx=A^Tb, " "with A(i,j) = 1/(2*n-(1+i+j)) and x(j) = 1"); + k = 0; + did_fail = CPL_FALSE; for (size = 1; size < nreps * nelem; size++) { cpl_matrix * p2; double error, residual; @@ -1595,6 +1598,13 @@ cpl_test_nonnull(product); xsolv = cpl_matrix_solve_normal(matrix, product); + if (cpl_error_get_code()) { + cpl_test_error(CPL_ERROR_SINGULAR_MATRIX); + cpl_test_null(xsolv); + did_fail = CPL_TRUE; + cpl_matrix_delete(product); + break; + } cpl_test_nonnull(xsolv); x_min = cpl_matrix_get_min(xsolv); @@ -1631,7 +1641,17 @@ "x_min=%g, x_max=%g", size, x_min, x_max); break; } + k++; + } + if (did_fail) { + /* Solving stopped prematurely. Normally, we should stop when the system + is near-singular and so ill-conditioned that the solution has an + element where not even the most significant bit is correct. For this + final iteration we will allow the solver to alternatively fail (due a + singular matrix). + This should add support for Debian sbuild on mips64el */ + cpl_test_leq(6, k); } cpl_msg_info("", "Compute the determinant of increasingly large "
