"Substitute" makes M2 crash

[signedout]

Hi,

I've ran into a situation where M2 crashes, and I don't know what to do about it. I've managed to isolate a fairly simply minimal non-working example:

K=frac(QQ[t]) 
R=K[y,x, MonomialOrder=>GLex]; 
use R; 
M=matrix{{x*(1/t^3),x*(1/(2*t))}} 
M=substitute(M, x=>1) 
substitute(M, t=>0) 
M=t*M 
substitute(M, t=>0) 

After running that, I get

-- SIGSEGV 
-* stack trace, pid: 32744 
0# 0x00005620E19031AD in /usr/bin/M2-binary 
1# 0x00005620E190338A in /usr/bin/M2-binary 
2# 0x00007FDCB28ADFD0 in /lib/x86_64-linux-gnu/libc.so.6 
3# 0x00005620E1AA31B8 in /usr/bin/M2-binary 
4# 0x00005620E1AA32F2 in /usr/bin/M2-binary 
5# 0x00005620E1AA3994 in /usr/bin/M2-binary 
6# 0x00005620E1A764C5 in /usr/bin/M2-binary 
7# 0x00005620E1A76A27 in /usr/bin/M2-binary 
8# 0x00005620E1A76F55 in /usr/bin/M2-binary 
9# 0x00005620E1A72C5B in /usr/bin/M2-binary 
10# 0x00005620E1AA2AF1 in /usr/bin/M2-binary 
11# 0x00005620E1AD829F in /usr/bin/M2-binary 
12# 0x00005620E1A73FBC in /usr/bin/M2-binary 
13# 0x00005620E1AE2626 in /usr/bin/M2-binary 
14# 0x00005620E1AD86CB in /usr/bin/M2-binary 
15# 0x00005620E1BDB5F6 in /usr/bin/M2-binary 
16# 0x00005620E195A44A in /usr/bin/M2-binary 
17# 0x00005620E192E7AC in /usr/bin/M2-binary 
18# 0x00005620E192F8FC in /usr/bin/M2-binary 
19# 0x00005620E193095D in /usr/bin/M2-binary 
20# 0x00005620E192EF5F in /usr/bin/M2-binary 
21# 0x00005620E192F868 in /usr/bin/M2-binary 
22# 0x00005620E193095D in /usr/bin/M2-binary 
23# 0x00005620E192EF5F in /usr/bin/M2-binary 
24# 0x00005620E192D752 in /usr/bin/M2-binary 
25# 0x00005620E1931260 in /usr/bin/M2-binary 
26# 0x00005620E192E7F1 in /usr/bin/M2-binary 
27# 0x00005620E192FF05 in /usr/bin/M2-binary 
28# 0x00005620E192EFB7 in /usr/bin/M2-binary 
29# 0x00005620E1933C0E in /usr/bin/M2-binary 
30# 0x00005620E1960AE8 in /usr/bin/M2-binary 
31# 0x00005620E1961195 in /usr/bin/M2-binary 
32# 0x00005620E19613FE in /usr/bin/M2-binary 
33# 0x00005620E192E7AC in /usr/bin/M2-binary 
34# 0x00005620E192E051 in /usr/bin/M2-binary 
35# 0x00005620E192D725 in /usr/bin/M2-binary 
36# 0x00005620E1933C0E in /usr/bin/M2-binary 
37# 0x00005620E1960AE8 in /usr/bin/M2-binary 
38# 0x00005620E1961470 in /usr/bin/M2-binary 
39# 0x00005620E1961EFB in /usr/bin/M2-binary 
40# 0x00005620E1902CF0 in /usr/bin/M2-binary 
41# 0x00005620E1BDFADB in /usr/bin/M2-binary 
42# 0x00005620E1BDFCBB in /usr/bin/M2-binary 
43# 0x00005620E1BE0259 in /usr/bin/M2-binary 
44# 0x00007FDCB394FE61 in /usr/lib/x86_64-linux-gnu/libgc.so.1 
45# 0x00007FDCB394AC42 in /usr/lib/x86_64-linux-gnu/libgc.so.1 
46# 0x00007FDCB350A6DB in /lib/x86_64-linux-gnu/libpthread.so.0 
47# 0x00007FDCB2990A3F in /lib/x86_64-linux-gnu/libc.so.6 
-- end stack trace *- 

[d-torrance]

I was able to reproduce this with debugging symbols:

i8 : substitute(M, t => 0)
-- SIGSEGV
-* stack trace, pid: 3487693
 0# stack_trace(std::ostream&, bool) at ../d/main.cpp:127
 1# segv_handler at ../d/main.cpp:244
 2# 0x00007F46693DB040 in /lib/x86_64-linux-gnu/libc.so.6
 3# FractionField::simplify(frac_elem*) const at ../e/frac.cpp:132
 4# FractionField::divide(ring_elem, ring_elem) const at ../e/frac.cpp:599
 5# PolyRing::imp_attempt_to_cancel_lead_term(ring_elem&, ring_elem, ring_elem&, int*) const at ../e/poly.cpp:1010
 6# PolyRing::remainderAndQuotient(ring_elem, ring_elem, ring_elem&) const at ../e/poly.cpp:1179
 7# PolyRing::divide(ring_elem, ring_elem) const at ../e/poly.cpp:986
 8# FractionField::eval(RingMap const*, ring_elem, int) const at ../e/frac.cpp:628
 9# RingMap::eval_term(Ring const*, ring_elem, int const*, int, int) const at ../e/ringmap.cpp:157
10# PolyRing::eval(RingMap const*, ring_elem, int) const at ../e/poly.cpp:1333
11# interface_rawRingMapEval at ../d/interface.dd:3328
12# evaluate_evalraw at ../d/evaluate.d:1297

[DanGrayson]

Here's what I get with a very recent version from the development branch, compiled in debug mode:

+ ./M2 --no-readline --print-width 169 --no-debug
Macaulay2, version 1.18.0.1
with packages: ConwayPolynomials, Elimination, IntegralClosure, InverseSystems, LLLBases, MinimalPrimes, PrimaryDecomposition, ReesAlgebra, Saturation, TangentCone

i1 : input "~/src/M2/M2.git/M2/BUILD/dan/bug14.m2"

ii2 : K=frac(QQ[t]) 

oo2 = K

oo2 : FractionField

ii3 : R=K[y,x, MonomialOrder=>GLex]; 

ii4 : use R; 

ii5 : M=matrix{{x*(1/t^3),x*(1/(2*t))}} 

oo5 = | 1/t3x 1/2tx |

              1       2
oo5 : Matrix R  <--- R

ii6 : M=substitute(M, x=>1) 

oo6 = | 1/t3 1/2t |

              1       2
oo6 : Matrix R  <--- R

ii7 : substitute(M, t=>0) 
../../bug14.m2:6:1:(3):[4]: error: division by zero!

[signedout]

Does this mean that this is a bug and if I want to get around it I should compile the latest development version?

[d-torrance]

The bug hasn't been fixed in the latest development version -- I reproduced it above in a very recent snapshot of the development branch. Dan used input, so it stopped at the first division by zero error. But the segfault is two lines later, after multiplying by t.

[DanGrayson]

Oh! I'll try running the subsequent lines.

[DanGrayson]

Okay, I reproduced the problem, and it appears that the engine code there doesn't anticipate the possibility of an error occurring in that bit of code. Mike can easily test the error flag and insert return statements, so I'm assigning this to him.

[signedout]

Many thanks to both of you for the quick replies. I hope it's not hard to fix!

[DanGrayson]

see #2178

[mikestillman]

Fixed in PR #2480.