Cadabra2 evaluate() crashed on segmentation fault

[mzavadil]

Cadabra2 command evaluate crashed on segmentation fault if a complete command was used for calculation:

Example:

# -*- coding: utf-8 -*-
from cadabra2 import *
print("start calculation...")
print(Coordinate(Ex(r"{r,t,f,v}"), Ex(r"")))
print(Indices(Ex(r"{a,b,c,d}"), Ex(r"values={t,r,f,v}, position=fixed")))
print(PartialDerivative(Ex(r"\partial{#}"), Ex(r"")))
print(Metric(Ex(r"g_{a b}"), Ex(r"")))
print(InverseMetric(Ex(r"g^{a b}"), Ex(r"")))
schwarzschild = Ex(r"{ g_{t t}=-(1-2*M/r),  g_{r r}=1/(1-2*M/r), g_{v v}=r**2, g_{f f}=r**2*sin(v)**2 }")
print(schwarzschild)
schwarzschild = complete(schwarzschild, Ex(r"g^{a b}"))
christoffel = Ex(r"\Gamma^{c}_{a b} = 1/2 g^{c d} ( \partial_{a}{g_{b d}} + \partial_{b}{g_{a d}} - \partial_{d}{g_{a b}} )")
print(christoffel)
connection = evaluate(christoffel, schwarzschild, rhsonly=True)
print(connection)
print("done.")

The problem is in dereference a null pointer of tree node of rhs part the expression in core/algorithms/substitute.cc.

Probably this is the cause of this error:

Traceback (most recent call last):
  File "cadabra2", line 248, in <module>
    exec(cmp)
  File "..\tests\crash3.py", line 12, in <module>
    print(schwarzschild)
RuntimeError: Found equals node with only one child node.

Occurred for printing a result of complete command:

...
print(complete(schwarzschild, Ex(r"g^{a b}")))
...

Tested on last windows build from commit: https://github.com/kpeeters/cadabra2/commit/5533cd5bcce492fbc5b9fafc551b08ee5aa19136

My version of Anaconda:

 active environment : base
    active env location : C:\tools\Anaconda3
            shell level : 1
       user config file : C:\Users\bigfo\.condarc
 populated config files : C:\Users\bigfo\.condarc
          conda version : 4.6.4
    conda-build version : 3.17.8
         python version : 3.7.2.final.0
       base environment : C:\Tools\Anaconda3  (writable)
           channel URLs : https://repo.anaconda.com/pkgs/main/win-64
                          https://repo.anaconda.com/pkgs/main/noarch
                          https://repo.anaconda.com/pkgs/free/win-64
                          https://repo.anaconda.com/pkgs/free/noarch
                          https://repo.anaconda.com/pkgs/r/win-64
                          https://repo.anaconda.com/pkgs/r/noarch
                          https://repo.anaconda.com/pkgs/msys2/win-64
                          https://repo.anaconda.com/pkgs/msys2/noarch
          package cache : C:\Tools\Anaconda3\pkgs
                          C:\Users\bigfo\.conda\pkgs
                          C:\Users\bigfo\AppData\Local\conda\conda\pkgs
       envs directories : C:\Tools\Anaconda3\envs
                          C:\Users\bigfo\.conda\envs
                          C:\Users\bigfo\AppData\Local\conda\conda\envs
               platform : win-64
             user-agent : conda/4.6.4 requests/2.21.0 CPython/3.7.2 Windows/10 Windows/10.0.17763
          administrator : False
             netrc file : None
           offline mode : False

and the sympy package:

sympy 1.3 py37_0
----------------
file name   : sympy-1.3-py37_0.tar.bz2
name        : sympy
version     : 1.3
build       : py37_0
build number: 0
size        : 9.5 MB
license     : 3-clause BSD
subdir      : win-64
url         : https://repo.anaconda.com/pkgs/main/win-64/sympy-1.3-py37_0.tar.bz2
md5         : 0f82e0de3a83b1e9e311621a603e9d2e
timestamp   : 2018-09-28 17:52:28 UTC
dependencies:
  - fastcache
  - mpmath >=0.19
  - python >=3.7,<3.8.0a0

[kpeeters]

Commit a9331e1fd9c0ead599d523f4c4322a67f2dbd065 messed up something which causes this (some self tests fail too). Haven't tracked it down yet.

[mzavadil]

The problem is may be in this change of core/SympyCdb.cc:

Return value of append_child was not used, but the call of it may be something do - just guess, I have not studied it in detail.

[kpeeters]

Yes, that one, and a similar one in core/algorithms/explicit_indices.cc. I went a bit too wild ripping out lines at every compiler warning. Now fixed in master.

[kpeeters]

By the way, in your code example, both complete and evaluate work directly on their argument and change it in place. There is no need to assign the result.

[mzavadil]

Thanks for a hint and quick fix.

If you want to eliminate every warnings, then here are some of the MS Build:

Build succeeded.

"C:\Users\bigfo\Workspace\dev\cadabra2\build\install.vcxproj" (default target) (1) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\ALL_BUILD.vcxproj" (default target) (3) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\tests\beginners_test.vcxproj" (default target) (4) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\core\cadabra2cadabra.vcxproj" (default target) (5) ->
(ClCompile target) ->
  c:\users\bigfo\workspace\dev\cadabra2\libs\whereami\whereami.c(49): warning C4005: 'WIN32_LEAN_AND_MEAN': macro redefinition [C:\Users\bigfo\Workspace\dev\cadabra2\build
\core\cadabra2cadabra.vcxproj]

"C:\Users\bigfo\Workspace\dev\cadabra2\build\install.vcxproj" (default target) (1) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\ALL_BUILD.vcxproj" (default target) (3) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\client_server\cadabra-server.vcxproj" (default target) (7) ->
  c:\users\bigfo\workspace\dev\cadabra2\libs\whereami\whereami.c(49): warning C4005: 'WIN32_LEAN_AND_MEAN': macro redefinition [C:\Users\bigfo\Workspace\dev\cadabra2\build
\client_server\cadabra-server.vcxproj]

"C:\Users\bigfo\Workspace\dev\cadabra2\build\install.vcxproj" (default target) (1) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\ALL_BUILD.vcxproj" (default target) (3) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\core\cadabra2.vcxproj" (default target) (8) ->
  c:\users\bigfo\workspace\dev\vcpkg\installed\x64-windows\include\pcrecpp.h(701): warning C4251: 'pcrecpp::RE::pattern_': class 'std::basic_string<char,std::char_traits<c
har>,std::allocator<char>>' needs to have dll-interface to be used by clients of class 'pcrecpp::RE' (compiling source file C:\Users\bigfo\Workspace\dev\cadabra2\core\Stor
age.cc) [C:\Users\bigfo\Workspace\dev\cadabra2\build\core\cadabra2.vcxproj]

"C:\Users\bigfo\Workspace\dev\cadabra2\build\install.vcxproj" (default target) (1) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\ALL_BUILD.vcxproj" (default target) (3) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\core\cadabra2.vcxproj" (default target) (8) ->
(Link target) ->
  LINK : warning LNK4075: ignoring '/INCREMENTAL' due to '/LTCG' specification [C:\Users\bigfo\Workspace\dev\cadabra2\build\core\cadabra2.vcxproj]

"C:\Users\bigfo\Workspace\dev\cadabra2\build\install.vcxproj" (default target) (1) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\ALL_BUILD.vcxproj" (default target) (3) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\frontend\gtkmm\cadabra2-gtk.vcxproj" (default target) (9) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\client_server\cadabra_client.vcxproj" (default target) (10) ->
(ClCompile target) ->
  c:\users\bigfo\workspace\dev\cadabra2\libs\whereami\whereami.c(49): warning C4005: 'WIN32_LEAN_AND_MEAN': macro redefinition [C:\Users\bigfo\Workspace\dev\cadabra2\build
\client_server\cadabra_client.vcxproj]

"C:\Users\bigfo\Workspace\dev\cadabra2\build\install.vcxproj" (default target) (1) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\ALL_BUILD.vcxproj" (default target) (3) ->
"C:\Users\bigfo\Workspace\dev\cadabra2\build\frontend\gtkmm\cadabra2-gtk.vcxproj" (default target) (9) ->
  c:\users\bigfo\workspace\dev\cadabra2\frontend\gtkmm\notebookwindow.cc(1795): warning C4805: '==': unsafe mix of type 'bool' and type 'int' in operation [C:\Users\bigfo\
Workspace\dev\cadabra2\build\frontend\gtkmm\cadabra2-gtk.vcxproj]
  c:\users\bigfo\workspace\dev\cadabra2\frontend\common\lodepng.cc(3411): warning C4334: '<<': result of 32-bit shift implicitly converted to 64 bits (was 64-bit shift int
ended?) [C:\Users\bigfo\Workspace\dev\cadabra2\build\frontend\gtkmm\cadabra2-gtk.vcxproj]

    7 Warning(s)
    0 Error(s)

Time Elapsed 00:03:44.17

[kpeeters]

Sorry, that's a bit too much for my restricted spare time. If you know how to fix it then a patch will be gratefully accepted, but otherwise I think this will just have to stay as it is.