test_intern in test_sys leaks negative references in free-threaded build

Bug report

The test_intern test "leaks" -2 references in the free-threaded build. This isn't caught by the refleak build bots because negative refleaks are treated as "suspicious" rather than "failing". This is a problem because the negative "leak" can hide a real leak somewhere else.

I think there's likely something wrong with the accounting for references in the immortalize=1 case in intern_common because this shows up in the free-threaded build, but not the default build.

./python -m test test_sys -m test_intern -R 3:3

Using random seed: 3390398091
0:00:00 load avg: 14.98 Run 1 test sequentially
0:00:00 load avg: 14.98 [1/1] test_sys
beginning 6 repetitions. Showing number of leaks (. for 0 or less, X for 10 or more)
123:456
X8. ...
test_sys leaked [-2, -2, -2] references, sum=-6 (this is fine)

Linked PRs

@encukou

cc @encukou

I think this is just an issue with refleak.py:

cpython/Lib/test/libregrtest/refleak.py

Lines 144 to 148 in 046670c

    
           interned_immortal_after = getunicodeinternedsize( 
        
               # Use an internal-only keyword argument that mypy doesn't know yet 
        
               _only_immortal=True)  # type: ignore[call-arg] 
        
           alloc_after = getallocatedblocks() - interned_immortal_after 
        
           rc_after = gettotalrefcount() - interned_immortal_after * 2

We now account for those two references in intern_common, so we shouldn't subtract them out in refleak.py:

cpython/Objects/unicodeobject.c

Lines 15518 to 15527 in 046670c

    
               if (!_Py_IsImmortal(s)) { 
        
                   /* The two references in interned dict (key and value) are not counted. 
        
                   unicode_dealloc() and _PyUnicode_ClearInterned() take care of this. */ 
        
                   Py_SET_REFCNT(s, Py_REFCNT(s) - 2); 
        
           #ifdef Py_REF_DEBUG 
        
                   /* let's be pedantic with the ref total */ 
        
                   _Py_DecRefTotal(_PyThreadState_GET()); 
        
                   _Py_DecRefTotal(_PyThreadState_GET()); 
        
           #endif 
        
               }

pythongh-122420: Fix accounting for immortal, interned strings in ref…

gh-122420: Fix accounting for immortal interned strings in refleak.py (…

pythongh-122420: Fix accounting for immortal interned strings in refl…

[3.13] gh-122420: Fix accounting for immortal interned strings in ref…

Navigation Menu

Repository navigation

`test_intern` in `test_sys` leaks negative references in free-threaded build #122420

Bug report

Linked PRs

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

	interned_immortal_after = getunicodeinternedsize(
	# Use an internal-only keyword argument that mypy doesn't know yet
	_only_immortal=True) # type: ignore[call-arg]
	alloc_after = getallocatedblocks() - interned_immortal_after
	rc_after = gettotalrefcount() - interned_immortal_after * 2

	if (!_Py_IsImmortal(s)) {
	/* The two references in interned dict (key and value) are not counted.
	unicode_dealloc() and _PyUnicode_ClearInterned() take care of this. */
	Py_SET_REFCNT(s, Py_REFCNT(s) - 2);
	#ifdef Py_REF_DEBUG
	/* let's be pedantic with the ref total */
	_Py_DecRefTotal(_PyThreadState_GET());
	_Py_DecRefTotal(_PyThreadState_GET());
	#endif
	}

test_intern in test_sys leaks negative references in free-threaded build #122420

Description

Bug report

Linked PRs

Activity

colesbury commented on Jul 29, 2024

colesbury commented on Jul 29, 2024

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions

`test_intern` in `test_sys` leaks negative references in free-threaded build #122420