[Coin-discuss] query re. optimised builds for COIN

[Lou Hafer]

Folks,

	As the last step in testing OsiDylp, I've been trying to do optimised
builds in various programming environments. I'm finding it very frustrating
and am wondering about the experiences of others. In all cases the debug
build runs correctly and has no runtime memory errors I'm aware of. Here's
what I've seen with optimised builds:

  * Sun Workshop 6 Update 1 on SPARC/Solaris 8 required rearrangment of code
    in OsiIndexedVector and OsiIndexedVectorTest in order to execute
    correctly. (The highlight was having CC claim that 1.0/1.0 == 24!)

  * GCC 2.95.2 on SPARC/Solaris 8 was happy with the same code as Sun
    Workshop.

  * GCC 2.96 on Intel/RedHat 7.1 dumps core in OsiPackedMatrix::getVector
    (gdb reports line 135 in OsiPackedMatrix.hpp). I haven't yet put serious
    time into debugging the problem.

I'm looking for answers to two specific questions:

  * I'd like to hear from people who have been able to make the optimised
    build with no problems. What was your programming environment? I'm
    assuming there are a fair number of you out there who have done this
    without problem.

  * Does anyone have experience with GCC 3.0.x? I've given thought to
    upgrading but have been discouraged by comments at the gcc web page (GCC
    3.0 caveats) about problems interacting with gdb. The caveats are dated
    June 2001, but I haven't seen anything that states the problems are
    fixed.


					Thanks,
						Lou