-------- Original Message -------- Subject: Re: Apparent non-termination of NTL's `make check' (version 5.3.2) Date: Sun, 4 Jul 2004 14:43:08 -0400 From: victor shoup shoup@cs.nyu.edu To: Roberto Bagnara bagnara@cs.unipr.it

Thanks...the code where it seems stuck looks pretty innocent. For now, I'll go on the assumption that it's a gcc bug. You should be able to safely use NTL, with LLL_FP.c compiled with -O1.

All the best, Victor

On Sunday, July 4, 2004, at 11:57 AM, Roberto Bagnara wrote:

victor shoup wrote:

...

I just thought of one other thing to try. Recompile everything with -O2, but ./configure with NTL_CLEAN_INT=on NTL_CLEAN_PTR=on To be honest, I'd be surprised if this worked, but if it does, that would be very interesting to me. These flags are supposed to avoid some code sequences that make assumptions that go beyond the language standard (but which seem to be universally implemented).

No, compiling with these flags does not eliminate the infinite loop.

...

Some other things to investigate: are the problems ONLY with the subset program?

Yes.

...

Finally, I guess one should try compiling with -g -O2, and run the subset program in gdb, force it to quit when its in an infinite loop, and get a backtrace...now that I think of it, I guess this is the first thing one should really try!

Yes, it seems to loop around here:

Program received signal SIGINT, Interrupt. BKZ_FP (BB=@0xbfffd740, UU=0x0, delta=0.98999999999999999, beta=20, prune=10, check=0x8049640 <SubsetSumSolution(NTL::vec_ZZ const&)>) at LLL_FP.c:1505 1505 for (i = jj; i <= kk; i++) { (gdb) info stack #0 BKZ_FP (BB=@0xbfffd740, UU=0x0, delta=0.98999999999999999, beta=20, prune=10, check=0x8049640 <SubsetSumSolution(NTL::vec_ZZ const&)>) at LLL_FP.c:1505 #1 0x08049fa1 in main () at subset.c:111 (gdb)

All the best,

Roberto

-- Prof. Roberto Bagnara Computer Science Group Department of Mathematics, University of Parma, Italy http://www.cs.unipr.it/~bagnara/ mailto:bagnara@cs.unipr.it