TestSuite: test that __new__ doesn't crash

[jdemeyer]

For any type t, the following should not crash Python:

t.__new__(t)

For Element classes, we also test

t.__new__(t, parent)

where parent is a parent (this mimics a typical call to __new__ for elements).

It is of course legal for this to raise ordinary exceptions.

One could argue that you should not call just __new__ like this since this will create invalid objects anyway. That is true, but this test checks for bugs in the object creation and deletion code. For example, if __new__ or __init__ or code calling __new__ raises a Python exception, then __dealloc__ should not segfault (which is exactly what caused the crash reported at #21628).

Bugs found: #21651, #21662, #21663, #21664

[jdemeyer]

New commits:

​74d05c0

Test that __new__ doesn't crash

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​d6eec0e

Test that __new__ doesn't crash

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​c0b3e39	Fix initialization of ComplexIntervalFieldElement
​2367f85	Fix initialization of rational vectors
​7323945	Fix initialization of integer vectors
​72d62c4	Allow singular_ring_delete(NULL)
​42c6423	Merge remote-tracking branches 'trac/u/jdemeyer/fix_initialization_of_complexintervalfieldelement', 'trac/u/jdemeyer/fix_initialization_of_vectors' and 'trac/u/jdemeyer/allow_singular_ring_delete_null_' into t/21658/testsuite_sageobject___test_that___new___doesn_t_crash_sage
​3880b48	Test that __new__ doesn't crash

[jdemeyer]

Test failures (every verbose run of TestSuite() needs to be patched).

[git]

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

​425533a	Merge tag '7.5.beta0' into t/21658/testsuite_sageobject___test_that___new___doesn_t_crash_sage
​8140d6f	Doctest fixes for _test_new() TestSuite test

[jdemeyer]

I only added some obvious doctest fixes, so I take the liberty to set this back to positive_review myself. I did run make ptestlong.