Fix ref count error issue #906 #907
Conversation
Codecov Report
@@ Coverage Diff @@
## master #907 +/- ##
==========================================
+ Coverage 72.59% 72.89% +0.29%
==========================================
Files 51 51
Lines 6474 6474
Branches 1302 1302
==========================================
+ Hits 4700 4719 +19
+ Misses 1378 1355 -23
- Partials 396 400 +4
Continue to review full report at Codecov.
|
|
Changes look good. I'll see if I can put together a regression test (but it might take me a few days to get around to it). |
…nt_callable_allow_none
|
I merged the branch for #910 (which contains a regression test) into this one. |
|
Looking at the code, there are some other issues here: for example the |
|
Have expanded the tests somewhat and streamlined the code a bit to test the quicker cases first. |
| @@ -2921,7 +2921,7 @@ trait_new(PyTypeObject *trait_type, PyObject *args, PyObject *kw) | |||
| } | |||
|
|
|||
| PyErr_Format( | |||
| TraitError, | |||
| TraitError, | |||
There was a problem hiding this comment.
Unrelated change: VS Code automatically trims trailing whitespace.
|
@midhun-pm I've made a bunch of changes on this PR; please could you review the current state after the changes? |
traits/tests/test_callable.py
Outdated
| string_value = "some string" | ||
| callable_value = my_function | ||
|
|
||
| for _ in range(10): |
There was a problem hiding this comment.
Hmm. No really good reason. Let me take out the loop and see if a single iteration is sufficient to cause the segfault.
There was a problem hiding this comment.
So a single iteration is not enough: on my own machine, I need at least 3. I've added a comment about the number of iterations.
My (unverified) conjecture is that what's happening is that on each iteration through the loop, my_function loses one of its references; when its reference count gets to zero, it gets garbage collected. Everything's fine unless we try to access my_function after it's been garbage collected. The second iteration is probably reducing the reference count to zero, and causing my_function to be garbage collected. The third iteration then accesses my_function and causes the segfault.
These changes look good. It's clean. Thanks! |
PR fixes issue #906