Update pynac to 0.7.27 to fix wrong result on symbolic exponentiation

[dkrenn]

Before this ticket:

sage: n((24*sqrt(3))^(100/50))
3.80392047155301e5927962146
sage: n((24*sqrt(3))^(2))
1728.00000000000

Clearly, both should be the same, namely the second.

This was on SageMath 9.1 on openSUSE Leap 15.2 (64bit) (reported by a colleague to me).

I've tried to reproduce, but my SageMath, as well as on CoCalc crashes. The error message on my machine was

Traceback (most recent call last):
  File "<string>", line 25, in <module>
ModuleNotFoundError: No module named 'Cython'
Error while executing Python code.
Saved trace to /home/dakrenn/.sage/crash_logs/crash_lqkqgydq.log
------------------------------------------------------------------------
Unhandled SIGABRT: An abort() occurred.
This probably occurred because a *compiled* module has a bug
in it and is not properly wrapped with sig_on(), sig_off().
Python will now terminate.
------------------------------------------------------------------------
Aborted (core dumped)

However, my SageMath seems to be fine otherwise, make ptestlong passes. Same on current 9.2.beta9.

[gh-DaveWitteMorris]

I confirm the error. I get exactly the same answers as in the original report. I am using Mac OS 10.15.5, and tested both 9.1 and 9.2b10. (The version with 100/2 is slow -- takes about 50 seconds.)

When I tried CoCalc, it failed because it ran out of memory. (The memory gauge was over 3GB before it aborted.)

[gh-DaveWitteMorris]

There is no need to use big numbers, so the situation is even worse than it first appeared:

sage: n((2*sqrt(2))^(2/1))                                                      
1.24131221754531e1292913987
sage: n((2*sqrt(2))^2)                                                          
8.00000000000000

This only takes a few seconds with 9.1 on CoCalc (but needs more than 1GB of memory) or with 9.2b13 on MacOS 10.15.6.

[mantepse]

I can confirm this on Ubuntu 18 with SageMath version 9.2.beta13, using Python 3.8.5.

[gh-bmlivin]

Okay, I believe I've tracked this down. It is a pynac bug. Here's what's happening:

Sage eventually calls pynac's pow from sage.symbolic.expression:Expression._pow_, but before that happens, it simplifies the exponent. Specifically, it simplifies the exponent so that pynac thinks it is an integer, but its type is still MPQ.

None of that should matter, except that what pynac calls to evaluate the expression is numeric::pow_intexp. Here is the content of that function:

if (not exponent.is_integer())
    throw std::runtime_error("nueric::pow_intexp: exponent not integer");
if (exponent.t == MPZ) {
    if (not mpz_fits_sint_p(exponent.v._bigint))
        throw std::runtime_error("size of exponent exceeds signed long size");
    return power(mpz_get_si(exponent.v._bigint));
}
return power(exponent.v._long);

In this function, exponent.is_integer() is True and exponent.t is MPQ. So both of the if statements are false, and power(exponent.v._long) is called. The problem is that v._long contains junk (very large junk!). Evaluating that statement takes a long time, and it won't give you the answer you want.

If, on the other hand, the second if statement were true, the result would have been exactly what you want. So simply checking whether exponent.t == MPQ or exponent.t == MPZ should fix this. I haven't tried this, but it's what I would at least try.

I think it may be possible as well to fix this by getting Sage to change the type of the exponent when it's cancelled out. I haven't looked into this and I think it sounds like a bad idea.

I'm unassigning this because I can't currently report it upstream.

[gh-bmlivin]

I've fixed this upstream and submitted a pull request: ​https://github.com/pynac/pynac/pull/355. The pull request also includes additions to the doctest for sage.symbolic.expression.Expression::_pow_.

[dimpase]

OK, trying ​https://patch-diff.githubusercontent.com/raw/pynac/pynac/pull/355.patch now.

But where are "additions to the doctest for sage.symbolic.expression.Expression::_pow_"?

I presume they should be here, on this ticket.

[dimpase]

OK, with that patch,

sage: QQ((24*sqrt(3))^(100/50))==1728                                                                                                                                
True

works. I'll add this as a doctest somewhere.

[dimpase]

patches from pynac PRs to be removed at the next pynac update.

---

New commits:

​dc073df

fixes for trac #30446

[gh-DaveWitteMorris]

This patch to pynac also fixes #28620, #30304, and #30786. The PR at #30786 adds doctests for all 3 of these tickets (and also moves the doctest of this ticket from the _latex_ method to the _pow_ method, which seems like a more appropriate place).

[gh-DaveWitteMorris]

fyi: Patchbot Linux/74-Ubuntu_SMP_Tue_Sep_17_17%3A06%3A04_UTC_2019/x86_64/4.15.0-65-generic/pc72 seems to have failed to build pynac at ticket #30786. ("Error installing package pynac-0.7.26.sage-2020-04-03.p1") If there is a genuine problem, I think it must be because of this ticket.

[gh-DaveWitteMorris]

This patch also seems to fix #31137. I added a doctest there.

[gh-DaveWitteMorris]

fyi: Another ubuntu patchbot failed to build pynac (on ticket #31137 this time). Both failures say WARNING: 'aclocal-1.16' is missing on your system. ... Makefile:432: recipe for target 'aclocal.m4' failed.

[vbraun]

Patch touches configure stuff which triggers automake run (which is not a Sage dependency)

Building pynac-0.7.26.sage-2020-04-03.p1
CDPATH="${ZSH_VERSION+.}:" && cd . && /bin/bash /var/lib/buildbot/slave/sage_git/build/local/var/tmp/sage/build/pynac-0.7.26.sage-2020-04-03.p1/src/missing aclocal-1.16 -I m4
/var/lib/buildbot/slave/sage_git/build/local/var/tmp/sage/build/pynac-0.7.26.sage-2020-04-03.p1/src/missing: line 81: aclocal-1.16: command not found
WARNING: 'aclocal-1.16' is missing on your system.
         You should only need it if you modified 'acinclude.m4' or
         'configure.ac' or m4 files included by 'configure.ac'.
         The 'aclocal' program is part of the GNU Automake package:
         <https://www.gnu.org/software/automake>
         It also requires GNU Autoconf, GNU m4 and Perl in order to run:
         <https://www.gnu.org/software/autoconf>
         <https://www.gnu.org/software/m4/>
         <https://www.perl.org/>
Makefile:432: recipe for target 'aclocal.m4' failed
make[5]: *** [aclocal.m4] Error 127
make[5]: Failed to remake makefile 'Makefile'.
make[5]: Target 'all' not remade because of errors.

[gh-mjungmath]

I set this to critical since many tickets (some already positively reviewed providing a doctest) rely on this one.

[mkoeppe]

​https://doc.sagemath.org/html/en/developer/packaging.html#when-to-patch-when-to-repackage-when-to-autoconfiscate explains that patching is not an option here. A pynac release needs to be made and the new release tarball needs to be used.

[dimpase]

I'll repurpose this ticket for the package update.

[dimpase]

New commits:

​9d747fc	sage --package update-latest: Distinguish pypi package name and spkg name
​f74f66c	sage --package update-latest: Accept package classes :standard:, :optional: etc., restrict to normal Python packages
​182b3d2	sage -package fix-checksum: Handle package classes, ignore non-normal packages
​9a57cf6	pynac update
​563940e	doctest from #30446

[mkoeppe]

When you make official releases in ​https://github.com/pynac/pynac, there is no need for a version suffix such as "sage-2020-04-03". I used this when I made the previous release from ​https://github.com/mkoeppe/pynac to make clear that this was an unofficial release.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​51def3c

sane version name

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​4834dc8

remove upstreamed patch

[dimpase]

should provide a cleaner tarball

[dimpase]

Replying to mkoeppe:

When you make official releases in ​https://github.com/pynac/pynac, there is no need for a version suffix such as "sage-2020-04-03". I used this when I made the previous release from ​https://github.com/mkoeppe/pynac to make clear that this was an unofficial release.

but then ​https://github.com/pynac/pynac/commit/bbaff9d21d31e73f74bf3dd751b16c4dd3ac28a7 has sneaked upstream in some way :-)

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​214712d

tarball update

[mkoeppe]

build/pkgs/pynac/SPKG.rst could use an update. In particular "Special Update/Build? Instructions" should be removed

[git]

Branch pushed to git repo; I updated commit sha1 and set ticket back to needs_review. New commits:

​ccdf77c

update SPKG.rst

[dimpase]

ok, done

[vbraun]

[dochtml] [calculus ] docstring of sage.symbolic.expression.Expression.subs:135: WARNING: Literal block expected; none found.
[dochtml] [calculus ] docstring of sage.symbolic.expression.Expression.substitute:135: WARNING: Literal block expected; none found.
[dochtml] [manifolds] The inventory files are in local/share/doc/sage/inventory/en/reference/manifolds.
[dochtml] Build finished. The built documents can be found in /home/release/Sage/local/share/doc/sage/inventory/en/reference/manifolds
[dochtml] [calculus ] The inventory files are in local/share/doc/sage/inventory/en/reference/calculus.
[dochtml] Error building the documentation.
[dochtml] Traceback (most recent call last):
[dochtml]   File "/usr/lib64/python3.9/runpy.py", line 197, in _run_module_as_main
[dochtml]     return _run_code(code, main_globals, None,
[dochtml]   File "/usr/lib64/python3.9/runpy.py", line 87, in _run_code
[dochtml]     exec(code, run_globals)
[dochtml]   File "/home/release/Sage/local/lib64/python3.9/site-packages/sage_setup/docbuild/__main__.py", line 2, in <module>
[dochtml]     main()
[dochtml]   File "/home/release/Sage/local/lib64/python3.9/site-packages/sage_setup/docbuild/__init__.py", line 1730, in main
[dochtml]     builder()
[dochtml]   File "/home/release/Sage/local/lib64/python3.9/site-packages/sage_setup/docbuild/__init__.py", line 343, in _wrapper
[dochtml]     getattr(get_builder(document), 'inventory')(*args, **kwds)
[dochtml]   File "/home/release/Sage/local/lib64/python3.9/site-packages/sage_setup/docbuild/__init__.py", line 569, in _wrapper
[dochtml]     self._build_everything_except_bibliography(lang, format, *args, **kwds)
[dochtml]   File "/home/release/Sage/local/lib64/python3.9/site-packages/sage_setup/docbuild/__init__.py", line 555, in _build_everything_except_bibliography
[dochtml]     build_many(build_ref_doc, non_references)
[dochtml]   File "/home/release/Sage/local/lib64/python3.9/site-packages/sage_setup/docbuild/__init__.py", line 295, in build_many
[dochtml]     _build_many(target, args, processes=NUM_THREADS)
[dochtml]   File "/home/release/Sage/local/lib64/python3.9/site-packages/sage_setup/docbuild/utils.py", line 289, in build_many
[dochtml]     raise worker_exc.original_exception
[dochtml] OSError: docstring of sage.symbolic.expression.Expression.subs:135: WARNING: Literal block expected; none found.
[dochtml] 
[dochtml]     Note: incremental documentation builds sometimes cause spurious
[dochtml]     error messages. To be certain that these are real errors, run
[dochtml]     "make doc-clean" first and try again.

[mkoeppe]

Could someone fix up the documentation markup please?

[gh-DaveWitteMorris]

I can't find any mistake, the patchbots are green, and the html documentation builds for me.

I wonder if maybe this error was posted to the wrong ticket? It is exactly the same error that was reported by a patchbot on ticket #30378 because I goofed up the WARNING block there:

[dochtml] OSError: docstring of sage.symbolic.expression.Expression.subs:135: WARNING: Literal block expected; none found.

[mkoeppe]

Thanks for checking! Works for me too (tested on top of 31344)