Project

General

Profile

Bug #1250

SEGV: CoCoALib with Normaliz in multithreading

Added by John Abbott about 5 years ago. Updated over 4 years ago.

Status:
Closed
Priority:
Urgent
Assignee:
Anna Maria Bigatti
Category:
Improving
Target version:
CoCoALib-0.99650 November 2019
Start date:
04 Mar 2019
Due date:
% Done:

100%

Estimated time:
2.11 h
Spent time:
2.10 h

Description

Bruns reports (email 2019-03-02) the following SEGV bug:

Thread 1 "normaliz" received signal SIGSEGV, Segmentation fault.
0x0000000000000013 in ?? ()
(gdb) bt
#0  0x0000000000000013 in ?? ()
#1  0x00000000008ce575 in CoCoA::SparsePolyRingBase::myMul(CoCoA::RingElemRawPtr, CoCoA::RingElemConstRawPtr, CoCoA::RingElemConstRawPtr) const ()
#2  0x00000000008a0945 in CoCoA::operator*(CoCoA::RingElemAlias const&, CoCoA::MachineInt const&) ()
#3  0x0000000000466968 in libnormaliz::VxM (V=..., M=...) at ../libnormaliz/nmz_polynomial.cpp:160
#4  0x00000000004779e0 in libnormaliz::substituteAndIntegrate (FF=..., A=..., degrees=..., R=..., Factorial=..., factQuot=..., lcmDegs=...)
    at ../libnormaliz/nmz_integral.cpp:106

Note It seems to be a problem with parallelization. If I switch to serial mode with -x=1, it disappears.

Note 2 The problem arises already with CoCoALib 0.99564. For the time being I return to 0.99562.

Related issues

Related to CoCoA-5 - Support #1240: John's visit Feb 2019Closed2019-02-08

History

#1 Updated by John Abbott about 5 years ago

  • Status changed from New to In Progress
  • % Done changed from 0 to 10

We are searching for suspect code. We are working blind (since we do not have any experience with parallel execution).

#2 Updated by Anna Maria Bigatti about 5 years ago

  • % Done changed from 10 to 20

Looking for possible problems in the logs from January to June 2018

date: 2018/04/20 18:51:25;  author: abbott;  state: Exp;  lines: +18 -3
Summary: Changed ctors for BigInt/BigRat from string or from MPZ/MPQ

#3 Updated by Anna Maria Bigatti about 5 years ago 

date: 2018/02/27 11:04:08;  author: abbott;  state: Exp;  lines: +6 -1
Summary: Added DetOverQQ

date: 2018/02/15 14:35:42;  author: abbott;  state: Exp;  lines: +21 -4
Summary: Added det4x4 and det5x5, DetByCRT

#4 Updated by Anna Maria Bigatti about 5 years ago

#5 Updated by Anna Maria Bigatti about 5 years ago

  • Assignee set to Anna Maria Bigatti
  • % Done changed from 20 to 90

The distribution had a precompiled PREPROCESSOR_DEFNS.H and autoconf.mk disabling CoCoA_THREADSAFE_HACK.
A partial configuration could cause inconsistent, but undetected, compilation flags.
I fixed the release script, and, for now, made a new release 0.99601 without the two files (which indeed should not be there!)

#6 Updated by John Abbott about 5 years ago

I have changed the configure script so that it prints out a clearer summary.
I have changed the Makefile so that veryclean remove all files created by configuration (except for last-config-cmd which is removed by distclean).

Hopefully, these changes should avoid a possible repeat of the situation.

#7 Updated by John Abbott about 5 years ago

  • Status changed from In Progress to Feedback

#8 Updated by John Abbott over 4 years ago

  • Status changed from Feedback to Closed
  • % Done changed from 90 to 100

This has been in "feedback" for 7 months, so I'm closing.

#9 Updated by Anna Maria Bigatti over 4 years ago

  • Estimated time set to 2.11 h

Also available in: Atom PDF