mikestillman/m2

The primary source code repository for Macaulay2

m2's Issues

EngineTests and ffpack

Should we test more ffpack fields?

Are we testing all of the ffpack linear algebra functions implemented in M2?

mpfr 3.0.1 and gmp 5.1.1 are not compatible

howto debug memory errors using valgrind+garbage collector

we should figure out, how to debug memory errors using valgrind+garbage collector,
or is it possible to disable gc?

(recall the bug in ColumnRankProfile, see
https://groups.google.com/forum/#!topic/ffpack-devel/2mwmhJpvQQ4
)
somewhat useful:
GC_DONT_GC environment flag

another helpful way is replacing GC_MALLOC with malloc
using LD_PRELOAD
e.g.

with malloc.c:

#include <stdio.h>
#include <stdlib.h>

void * GC_malloc(size_t a)
{
 // printf("calling my malloc\n");
  return malloc(a);
}

gcc -fPIC -shared malloc.c -o libmygc.so

export LD_PRELOAD = ./libmygc.so

check flint issue related to interchangeability of mpir and gmp

configuring and building flint succeeds, but
flint seems to link with gmp rather than with mpir and the check fails.

• can you confirm this issue?

I looked at the generated Makefile which always seems to contain lgmp

I suggest that at least if we configure M2 with --enable-build-libraries=mpir,
flint gets passed configure option --with-mpir
and if that is the case, flint will link with MPIR
( I have a patch for flint's configure file)

• comments/suggestions?
• Could we ask Dan for suggestions/help here?

testNorm fails from time to time for complex numbers

testNorm(CC_53) -- fails from time to time
testNorm(CC_100) -- fails from time to time
testNorm(CC_200) -- fails from time to time
testNorm(CC_54) -- fails from time to time

here is the minimal CC_53 failing setup :

entry := toCC( .81662363033492791p53, .97905576575740838p53 )

MM := mutableMatrix {{ entry }}

a := norm MM
b := norm matrix M
ans := (flatten entries MM)/abs//max;
a==b --ok
a==ans --false

((flatten entries MM)/abs)#0 ==a --false

if the failure is not reproducible, I will provide more information.

factory bugs and issues

Collect my list of bugs for factorization and characteristic sets to Jakob. He can try them on the latest devel versions of that library.

jk: put tests into Macaulay2/tests ?

fillMatrix fails for FFPACK with char=2 (matrix is always zero)

R = ZZpFFPACK( 2 );
M = mutableMatrix(R,10,10)

while (zero M ) do (fillMatrix M ) --infinite loop; M always zero...

needs some debugging.

Flint upstream report

Hi Mike,

I just completed an upstream report related to out Flint patch:
flintlib/flint#43
flintlib/flint#44

Are some of our/your remaining changes to Flint worthwhile for an upstream report?

execstack command for M2 binary does not work for SL 6.4

have no clue why.
Workaround:

add '-' at the start of correspionding line in
Macaulay2/bin/Makefile

build will fail when enabling givaro but not fflas-ffpack

using older mpfr3.0.1

if you want to use the older mpfr 3.0.1
in the linalg branch

you have to replace the
mpz_clears() call in
M2/Macaulay2/e/cra.cpp
by single mpz_clear calls:

// mpz_clears(a1,a2,u1,u2,q,h,mhalf,u2sqr,a2sqr,(void *)0);
mpz_clear(a1);
mpz_clear(a2);
mpz_clear(u1);
mpz_clear(u2);
mpz_clear(q);
mpz_clear(h);
mpz_clear(mhalf);
mpz_clear(u2sqr);
mpz_clear(a2sqr);