Remove Python access from Parent._element_constructor

[jdemeyer]

In order to avoid confusion between _element_constructor_ (a public interface) and _element_constructor (a private attribute), we change the latter from cdef public to cdef. The same for _element_init_pass_parent. This way, those attributes can no longer be accessed from Python.

Depends on #23882 Depends on #23884 Depends on #23894 Depends on #23899 Depends on #23905 Depends on #23907 Depends on #23914 Depends on #24033

CC: @simon-king-jena @koffie

Component: coercion

Author: Jeroen Demeyer, Simon King

Branch: 41614a7

Reviewer: Travis Scrimshaw

Issue created by migration from https://trac.sagemath.org/ticket/23881

[jdemeyer]

Changed branch from u/SimonKing/rename_parentelementconstructor____parentconstruct_element to u/jdemeyer/rename_parentelementconstructor____parentconstruct_element

[jdemeyer]

Changed commit from c95b724 to 41614a7

[jdemeyer]

New commits:

6878e67	Avoid _element_constructor in padics
44bc9fb	Do not assign _element_constructor in combinat/sf
f684cb0	Implement `_element_constructor_` for Homset
8b9765e	Avoid _element_constructor in geometry doctests
2630849	Various minor changes to _element_constructor
2856051	Merge commit '6878e67e99f1f3707c8b4a060264f5484989c52b'; commit '44bc9fb660fe4508b0bd8340667b8b65b70226a1'; commit 'f684cb0e4834a0041da29ef7a1f8f0b01cf84994'; commit '8b9765e781f8f9359cfc6021f6cd3d0e3b4b36d6'; commit '2630849a988376b44769460989a41919bbe89520' into t/23881/rename_parent__element_constructor____parent___construct_element
77ccb85	Do not set _element_init_pass_parent
41614a7	Remove Python access from Parent._element_constructor and Parent._element_init_pass_parent

[tscrim]

comment:51

Green patchbot => positive review.

[tscrim]

Reviewer: Travis Scrimshaw

[koffie]

comment:52

Both arando and rk02-math seem to fail the same test:

 sage -t --long src/sage/schemes/elliptic_curves/sha_tate.py # 1 doctest failed 

Which they do not seem to fail on with other tickets.

[tscrim]

comment:53

I don't see how those failures could be related to the changes in this ticket unless they appear also in one of the dependencies. Even still, I don't see how they could be connected, much less be machine dependent.

[koffie]

comment:54

I don't see see why either so I agree it is highly unlikely, but since parent is modified it is not impossible. It does not need to be machine dependent, it can also be a bug that happens 50% of the time that is exposed after this.

There are two machines failing on this ticket and I have not seen it on either arando or rk02-math in at least 10 other failing logs that I checked. So from this point of view it is also very unlikely that it is not from this ticket. Of course this ticket could have been very unlucky to expose a rare random bug twice for the first time but this is also highly unlikely.

I just would like to see arguments for why we are in one unlikely scenario and not the other.

Note that this ticket also has a lot of dependencies merged so it could also be exposed by one of the dependencies, which might explain why I didn't see it on any of the arando and rk02-math logsI checked.

If only we could search through the patchbot log easily, then we could get conclusive answers much faster.

[jdemeyer]

comment:55

I don't think that the failures have anything to do with this ticket, so I'm setting it to positive_review. If the failures are real, then it will fail on Volker's buildbot anyway.

[jdemeyer]

comment:56

I also tested it on arando manually (without the patchbot) and that test passed.

[koffie]

comment:57

Before you reported that arando passes all test I kicked the patchbot to get more info. And arando did fail again on the run resulting from that kick, while the bug did not occur in two tests ran on arando in between.

I am a bit mystified by the situation since it can't be the optional packages since rk02-math has none, so it must be a randomly occuring bug.

[koffie]

comment:58

I tested this with

./sage -t --long --global-iterations=100 src/sage/schemes/elliptic_curves/sha_tate.py

on arando and they all pass so it is not easily reproducible.

[koffie]

comment:59

The sha_tate doctest fails on sage8.1.beta7 without any patches on arando so I created #23962, since arando failed on the 3 tests it ran on this ticket, but none of the other beta6 tests, I still suspect it to be something introduced by one of the dependencies of this ticket that got merged in sage8.1.beta7. Anyway it can't be this ticket so lets move the discussion to #23962.

[jdemeyer]

comment:60

Replying to @tscrim:

to prevent anyone else from adding something (although I would say that is highly unlikely).

It seems that precisely this has happened in #23660...

[jdemeyer]

Changed dependencies from #23882, #23884, #23894, #23899, #23905, #23907, #23914 to #23882, #23884, #23894, #23899, #23905, #23907, #23914, #24033

[vbraun]

Changed branch from u/jdemeyer/rename_parentelementconstructor____parentconstruct_element to 41614a7

[jdemeyer]

comment:63

Good that this is merged now...

[jdemeyer]

Changed commit from 41614a7 to none