[ADOL-C] Fwd: sparse hessian: Conditional jump or move depends on uninitialised value(s)

[João Leal]

Hello Prof. Dr. Walther,

Thank you for the reply! "option[0] = 2" does fix the problem.

I ran the program with valgrind and it also reported multiple lines, such
as:

==6462== Conditional jump or move depends on uninitialised value(s)
==6462==    at 0x4EDA8C1: traverse_unary (in
/usr/lib/x86_64-linux-gnu/libadolc.so.2.1.0)
==6462==    by 0x4EDA8B0: traverse_unary (in
/usr/lib/x86_64-linux-gnu/libadolc.so.2.1.0)
==6462==    by 0x4ED9754: nonl_ind_forward_safe (in
/usr/lib/x86_64-linux-gnu/libadolc.so.2.1.0)
==6462==    by 0x4EF1696: sparse_hess (in
/usr/lib/x86_64-linux-gnu/libadolc.so.2.1.0)
==6462==    by 0x4018DB: main (adolc_mem.cpp:61)
==6462==  Uninitialised value was created by a heap allocation
==6462==    at 0x4C2BBCF: malloc (in
/usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==6462==    by 0x4ED94C0: nonl_ind_forward_safe (in
/usr/lib/x86_64-linux-gnu/libadolc.so.2.1.0)
==6462==    by 0x4EF1696: sparse_hess (in
/usr/lib/x86_64-linux-gnu/libadolc.so.2.1.0)
==6462==    by 0x4018DB: main (adolc_mem.cpp:61)

but it does not run out of memory (outside valgrind).

"option[0] = 1" will also cause it to run out of memory in the first run.
Since the problem appears to be related with sparsity pattern evaluation,
using option[0] = 0 or 1 with repeat = 1 after the first run does not run
out of memory.


Best regards,

João Leal
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/adol-c/attachments/20160711/e73f796d/attachment.html>