Upgrade to pari 2.17.1, cypari 2.2.1, cysignals 1.12.3 #38749

antonio-rojas · 2024-10-01T17:12:20Z

NEXT_PRIME_VIADIFF is removed in 2.17, port pari_prime_range to pari_PRIMES instead

Needs sagemath/cypari2#165 applied to cypari

antonio-rojas · 2024-10-01T17:14:07Z

Needs work because this is not compatible with 2.15. So either it needs to be merged with the pari upgrade or be made to work with older pari too (no idea how).

Running tests now.

antonio-rojas · 2024-10-03T07:43:45Z

Many test failures.

jhpalmieri · 2024-10-05T20:58:48Z

If the changes here eventually allow us to use a system Pari 2.17, then we should undo the change in #38772.

antonio-rojas · 2024-10-06T11:27:40Z

With this MR and the cypari fixes sagemath/cypari2#165 and sagemath/cypari2#166 all tests are passing with pari 2.17.

The changes are not compatible with 2.15 though, making them compatible requires more work. Also, some pari opeations (such as the number field prime ideals above a given prime) give random output with 2.17, which makes it harder to test. To solve both issues (and make tests more future proof), we should gradually move away from testing the exact output to just testing that the output is correct.

src/sage/schemes/elliptic_curves/ell_point.py

tobiasdiez · 2025-01-28T12:40:20Z

I don't want to spoil the happiness, but it looks like on Python 3.12 there are a couple of failed tests (that I've never seen before failing): https://github.com/sagemath/sage/actions/runs/13007996602/job/36279129460?pr=38749#step:10:5594

It might be only one of these temporary issues, maybe an admin can restart the failing run to see if it persists.

user202729 · 2025-01-28T13:10:41Z

For reference:

File "src/sage/matroids/transversal_matroid.pyx", line 701, in sage.matroids.transversal_matroid.TransversalMatroid.transversal_extension
Failed example:
    M2 = M.transversal_extension(element='d', newset=True)
Exception raised:
    Traceback (most recent call last):
      File "/home/runner/miniconda3/envs/sage-dev/lib/python3.12/site-packages/sage/doctest/forker.py", line 728, in _run
        self.compile_and_execute(example, compiler, test.globs)
      File "/home/runner/miniconda3/envs/sage-dev/lib/python3.12/site-packages/sage/doctest/forker.py", line 1152, in compile_and_execute
        exec(compiled, globs)
      File "<doctest sage.matroids.transversal_matroid.TransversalMatroid.transversal_extension[3]>", line 1, in <module>
        M2 = M.transversal_extension(element='d', newset=True)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "transversal_matroid.pyx", line 671, in sage.matroids.transversal_matroid.TransversalMatroid.transversal_extension
      File "transversal_matroid.pyx", line 792, in sage.matroids.transversal_matroid.TransversalMatroid.transversal_extension
      File "transversal_matroid.pyx", line 181, in sage.matroids.transversal_matroid.TransversalMatroid.__init__
    ValueError: set labels cannot be element labels

Doesn't look temporary but that part of code seem to have nothing to do with pari either.

collares · 2025-01-28T13:40:16Z

I don't want to spoil the happiness, but it looks like on Python 3.12 there are a couple of failed tests (that I've never seen before failing): https://github.com/sagemath/sage/actions/runs/13007996602/job/36279129460?pr=38749#step:10:5594

You haven't seen it before because the tests were introduced in 10.6.beta5 (#39155). But current master is also failing on the same tests: see, e.g., https://github.com/sagemath/sage/actions/runs/12979684201/job/36195720098

`NEXT_PRIME_VIADIFF` is removed in 2.17, port `pari_prime_range` to `pari_PRIMES` instead Needs sagemath/cypari2#165 applied to cypari URL: sagemath#38749 Reported by: Antonio Rojas Reviewer(s): Antonio Rojas, Dima Pasechnik, Gonzalo Tornaría

collares · 2025-01-29T18:34:16Z

Probably unrelated, but while testing this PR I hit a timeout in src/sage/groups/perm_gps/partn_ref/refinement_matrices.pyx (arguments --long --random-seed=148318620575155482011548963229209273971). Can anyone reproduce?

collares · 2025-01-29T18:58:10Z

Probably unrelated, but while testing this PR I hit a timeout in src/sage/groups/perm_gps/partn_ref/refinement_matrices.pyx (arguments --long --random-seed=148318620575155482011548963229209273971). Can anyone reproduce?

I can also reproduce the test hang with PARI 2.15, so this PR is not the problem. Filed #39407.

`NEXT_PRIME_VIADIFF` is removed in 2.17, port `pari_prime_range` to `pari_PRIMES` instead Needs sagemath/cypari2#165 applied to cypari URL: sagemath#38749 Reported by: Antonio Rojas Reviewer(s): Antonio Rojas, Dima Pasechnik, Gonzalo Tornaría

vbraun · 2025-02-09T12:24:09Z

On macos (mac mini m2) I'm getting

[cysignals-1.12.3] [spkg-install] The Meson build system
[cysignals-1.12.3] [spkg-install] Version: 1.3.1
[cysignals-1.12.3] [spkg-install] Source dir: /Users/buildbot-sage/worker/sage_git/build/local/var/lib/sage/venv-python3.12.5/var/tmp/sage/build/cysignals-1.12.3/src
[cysignals-1.12.3] [spkg-install] Build dir: /Users/buildbot-sage/worker/sage_git/build/local/var/lib/sage/venv-python3.12.5/var/tmp/sage/build/cysignals-1.12.3/src/.mesonpy-oe6svlmq
[cysignals-1.12.3] [spkg-install] Build type: native build
[cysignals-1.12.3] [spkg-install] Project name: cysignals
[cysignals-1.12.3] [spkg-install] Project version: undefined
[cysignals-1.12.3] [spkg-install] C compiler for the host machine: gcc (clang 15.0.0 "Apple clang version 15.0.0 (clang-1500.3.9.4)")
[cysignals-1.12.3] [spkg-install] C linker for the host machine: gcc ld64 classic  PROJECT:ld64
[cysignals-1.12.3] [spkg-install] 
[cysignals-1.12.3] [spkg-install] ../meson.build:1:0: ERROR: Unable to get clang pre-processor defines:
[cysignals-1.12.3] [spkg-install] Compiler stdout:
[cysignals-1.12.3] [spkg-install] 
[cysignals-1.12.3] [spkg-install] -----
[cysignals-1.12.3] [spkg-install] Compiler stderr:
[cysignals-1.12.3] [spkg-install] error: invalid argument '-std=gnu++11' not allowed with 'C'
[cysignals-1.12.3] [spkg-install] 
[cysignals-1.12.3] [spkg-install] -----

`NEXT_PRIME_VIADIFF` is removed in 2.17, port `pari_prime_range` to `pari_PRIMES` instead Needs sagemath/cypari2#165 applied to cypari URL: sagemath#38749 Reported by: Antonio Rojas Reviewer(s): Antonio Rojas, Dima Pasechnik, Gonzalo Tornaría

dimpase · 2025-02-09T15:15:25Z

clang 15? This is a very old Xcode, can you upgrade?

vbraun · 2025-02-09T15:20:45Z

Thats XCode 15.4 released in May 2024

vbraun@Mini-M2 ~ % gcc --version
Apple clang version 15.0.0 (clang-1500.3.9.4)
Target: arm64-apple-darwin23.6.0
Thread model: posix
InstalledDir: /Applications/Xcode-15.4.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin

dimpase · 2025-02-09T15:40:26Z

and no update available?

dimpase · 2025-02-09T15:59:35Z

I tested on William's M1 box:

dima@studio ~ % gcc -v
Apple clang version 16.0.0 (clang-1600.0.26.4)
Target: arm64-apple-darwin24.1.0
Thread model: posix
InstalledDir: /Library/Developer/CommandLineTools/usr/bin
dima@studio ~ % uname -a
Darwin studio.local 24.1.0 Darwin Kernel Version 24.1.0: Thu Oct 10 21:03:15 PDT 2024; root:xnu-11215.41.3~2/RELEASE_ARM64_T6000 arm64

and all was good.

dimpase · 2025-02-09T20:50:40Z

@vbraun - on the other hand it's a bug of sorts, in cysignals, most probably. @tobiasdiez - here cysignals meson build fails due to an over-strict C compiler which errors upon finding a C++-only flag.

vbraun · 2025-02-10T00:18:41Z

There is some unrelated bug that keeps me from updating, but also we should build on a less than one year old compiler release...

jhpalmieri · 2025-02-10T01:02:29Z

I see this on OS X:

sage -t --long --warn-long 30.0 --random-seed=73715025106880345904056137452377696661 src/sage/schemes/elliptic_curves/gp_simon.py
**********************************************************************
File "src/sage/schemes/elliptic_curves/gp_simon.py", line 84, in sage.schemes.elliptic_curves.gp_simon.?
Failed example:
    E.simon_two_descent()                             # long time
Expected:
    (0, 2, [])
Got:
    (2, 2, [(-121/64*w - 2719/64 : -15555/512*w - 186181/512 : 1)])

For what it's worth:

I don't get this failure with a different random seed.
With the develop branch (using Sage's pari 2.15.5), I get a failure but a different one if I use this random seed.
So this may be a pre-existing bug in a new guise.

user202729 · 2025-02-10T01:09:48Z

@jhpalmieri Looks like that's an improvement instead of a bug? The first and second number are the lower and upper bound of the rank of MW group, and the output seems to mean in newer PARI version, the code can successfully prove that the rank is exactly 2 instead of just "between 0 and 2".

Related: #9322

But because the test is intended to check that the function call doesn't take too long, maybe we should change it to a harder test case?

@JohnCremona any comment on this?

JohnCremona · 2025-02-10T10:10:58Z

@jhpalmieri Looks like that's an improvement instead of a bug? The first and second number are the lower and upper bound of the rank of MW group, and the output seems to mean in newer PARI version, the code can successfully prove that the rank is exactly 2 instead of just "between 0 and 2".

Related: #9322

But because the test is intended to check that the function call doesn't take too long, maybe we should change it to a harder test case?

@JohnCremona any comment on this?

It is not quite as wonderful as you suggest @user202729 . In the old version we only learn that the rank is bounded by 2, with no lower bound and no points. In the new version although the output claims that the rank is 2, it only gives one point, and in fact it has only proved that the lower bound is 1. But conjecturally the difference between the actual rank and the bound (which is called the 2-Se;mer rank) is even (this is a consequence of part of the Birch Swinnerton-Dyer conjectures, and is called the 'parity conjecture'), and Simon's script increases the lower bound by one in such cases.

Unless someone can point me to a specific result towards the parity conjecture (there are results in this direction) which justifies this, I would only rigorously conclude from the new output that the rank is either 1 or 2, however much I believe that the rank must in fact be 2, as I would need to be shown a second independent point to be convinced.

When I first wrote a Sage wrapper around Simon's script, years ago, I was careful not to use the parity conjecture, so that the lower bound was always equal to the number of points of initinite order returned. (Older versions of Simon's GP code were also a bit careless in the list of points, which might include points of finite order (which do not ontribute to the rank).)

In this case, all we are reporting is what the script outputs, so just change the expected output to the actual output. Perhaps with a comment saying that the lower bound has been increased by 1 using the parity conjecture.

I did think that now that lipari itself has a very good implementation of 2-descent, we had abandoned Simon's GP version (after many years of useful service). But that is only for curves defined over Q.

saraedum · 2025-02-10T16:04:55Z

cc @AndreasEnge who is waiting for this to package for guix.

dimpase · 2025-02-10T16:42:33Z

There is some unrelated bug that keeps me from updating, but also we should build on a less than one year old compiler release...

IIRC correctly, Apple's clang 15 started to convert quite harmless warnings into errors, something that was then rolled back.
One can in principle try to pass an appropriate option in CFLAGS, or sanitise CFLAGS...

tobiasdiez · 2025-02-10T20:48:25Z

On macos (mac mini m2) I'm getting

[cysignals-1.12.3] [spkg-install] The Meson build system
[cysignals-1.12.3] [spkg-install] Version: 1.3.1
[cysignals-1.12.3] [spkg-install] Source dir: /Users/buildbot-sage/worker/sage_git/build/local/var/lib/sage/venv-python3.12.5/var/tmp/sage/build/cysignals-1.12.3/src
[cysignals-1.12.3] [spkg-install] Build dir: /Users/buildbot-sage/worker/sage_git/build/local/var/lib/sage/venv-python3.12.5/var/tmp/sage/build/cysignals-1.12.3/src/.mesonpy-oe6svlmq
[cysignals-1.12.3] [spkg-install] Build type: native build
[cysignals-1.12.3] [spkg-install] Project name: cysignals
[cysignals-1.12.3] [spkg-install] Project version: undefined
[cysignals-1.12.3] [spkg-install] C compiler for the host machine: gcc (clang 15.0.0 "Apple clang version 15.0.0 (clang-1500.3.9.4)")
[cysignals-1.12.3] [spkg-install] C linker for the host machine: gcc ld64 classic  PROJECT:ld64
[cysignals-1.12.3] [spkg-install] 
[cysignals-1.12.3] [spkg-install] ../meson.build:1:0: ERROR: Unable to get clang pre-processor defines:
[cysignals-1.12.3] [spkg-install] Compiler stdout:
[cysignals-1.12.3] [spkg-install] 
[cysignals-1.12.3] [spkg-install] -----
[cysignals-1.12.3] [spkg-install] Compiler stderr:
[cysignals-1.12.3] [spkg-install] error: invalid argument '-std=gnu++11' not allowed with 'C'
[cysignals-1.12.3] [spkg-install] 
[cysignals-1.12.3] [spkg-install] -----

This should be fixed in more recent versions of meson, eg with mesonbuild/meson#13289.

dimpase · 2025-02-10T21:55:15Z

is this fix already in a released version of meson?
Then we should upgrade

github-actions bot added the s: needs review label Oct 1, 2024

antonio-rojas added s: needs work and removed s: needs review labels Oct 1, 2024

github-actions bot added s: needs review and removed s: needs work labels Oct 2, 2024

antonio-rojas changed the title ~~Fix build with pari 2.17~~ [WIP] Fix build with pari 2.17 Oct 3, 2024

antonio-rojas changed the title ~~[WIP] Fix build with pari 2.17~~ [WIP] Fix build and tests with pari 2.17 Oct 3, 2024

jhpalmieri mentioned this pull request Oct 5, 2024

external Pari 2.17 does not work (yet) #38769

Closed

2 tasks

antonio-rojas force-pushed the pari-2.17 branch from a6acbfb to f619b6f Compare October 5, 2024 20:01

antonio-rojas changed the title ~~[WIP] Fix build and tests with pari 2.17~~ Fix build and tests with pari 2.17 Oct 6, 2024

antonio-rojas force-pushed the pari-2.17 branch from 257246c to b3c57e1 Compare October 12, 2024 09:59

antonio-rojas added s: needs work and removed s: needs review labels Oct 12, 2024

github-actions bot added s: needs review and removed s: needs work labels Oct 12, 2024

antonio-rojas added s: needs work and removed s: needs review labels Oct 16, 2024

collares mentioned this pull request Oct 22, 2024

pari: 2.15.5 -> 2.17.1 NixOS/nixpkgs#348442

Merged

github-actions bot added s: needs review and removed s: needs work labels Oct 27, 2024

antonio-rojas added s: needs work and removed s: needs review labels Oct 27, 2024

tornaria reviewed Oct 30, 2024

View reviewed changes

src/sage/schemes/elliptic_curves/ell_point.py Outdated Show resolved Hide resolved

tornaria reviewed Oct 30, 2024

View reviewed changes

src/sage/schemes/elliptic_curves/ell_point.py Outdated Show resolved Hide resolved

github-actions bot added s: needs review and removed s: needs work labels Oct 30, 2024

tobiasdiez mentioned this pull request Jan 28, 2025

Salvage work on gammoids #39155

Merged

tobiasdiez mentioned this pull request Feb 6, 2025

Add support for Python 3.13 in conda #39189

Draft

5 tasks

vbraun added the s: needs work label Feb 10, 2025

dimpase mentioned this pull request Feb 10, 2025

upgrade meson to 1.7.0 #39490

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Upgrade to pari 2.17.1, cypari 2.2.1, cysignals 1.12.3 #38749

Upgrade to pari 2.17.1, cypari 2.2.1, cysignals 1.12.3 #38749

antonio-rojas commented Oct 1, 2024

antonio-rojas commented Oct 1, 2024

antonio-rojas commented Oct 3, 2024

jhpalmieri commented Oct 5, 2024

antonio-rojas commented Oct 6, 2024 •

edited

Loading

tobiasdiez commented Jan 28, 2025 •

edited

Loading

user202729 commented Jan 28, 2025

collares commented Jan 28, 2025

collares commented Jan 29, 2025

collares commented Jan 29, 2025

vbraun commented Feb 9, 2025

dimpase commented Feb 9, 2025

vbraun commented Feb 9, 2025

dimpase commented Feb 9, 2025

dimpase commented Feb 9, 2025

dimpase commented Feb 9, 2025

vbraun commented Feb 10, 2025

jhpalmieri commented Feb 10, 2025

user202729 commented Feb 10, 2025

JohnCremona commented Feb 10, 2025

saraedum commented Feb 10, 2025

dimpase commented Feb 10, 2025

tobiasdiez commented Feb 10, 2025

dimpase commented Feb 10, 2025

Upgrade to pari 2.17.1, cypari 2.2.1, cysignals 1.12.3 #38749

Are you sure you want to change the base?

Upgrade to pari 2.17.1, cypari 2.2.1, cysignals 1.12.3 #38749

Conversation

antonio-rojas commented Oct 1, 2024

antonio-rojas commented Oct 1, 2024

antonio-rojas commented Oct 3, 2024

jhpalmieri commented Oct 5, 2024

antonio-rojas commented Oct 6, 2024 • edited Loading

tobiasdiez commented Jan 28, 2025 • edited Loading

user202729 commented Jan 28, 2025

collares commented Jan 28, 2025

collares commented Jan 29, 2025

collares commented Jan 29, 2025

vbraun commented Feb 9, 2025

dimpase commented Feb 9, 2025

vbraun commented Feb 9, 2025

dimpase commented Feb 9, 2025

dimpase commented Feb 9, 2025

dimpase commented Feb 9, 2025

vbraun commented Feb 10, 2025

jhpalmieri commented Feb 10, 2025

user202729 commented Feb 10, 2025

JohnCremona commented Feb 10, 2025

saraedum commented Feb 10, 2025

dimpase commented Feb 10, 2025

tobiasdiez commented Feb 10, 2025

dimpase commented Feb 10, 2025

antonio-rojas commented Oct 6, 2024 •

edited

Loading

tobiasdiez commented Jan 28, 2025 •

edited

Loading