Bug report

There is other bug similar to#143378 in the Python implementation of BytesIO.

The buffer is exported twice for the input data. First, to determine the number of bytes, so we can create a slice when overwriting BytesIO data after seeking. Second, to get the data itself. If the number of bytes different, the underlying bytearray object can be shrinked or expanded, withmoving bytes after the current position.

Solution -- we need to use the same exported bufferwith memoryview() for the whole writing operation. The test will be slightly more complicated variant of the tests added in#143408.

@superboy-zjc, are you interesting to take this issue? It should be easy to you.

Linked PRs

• gh-143602: Fix duplicate buffer exports in io.BytesIO.write #143629
• [3.14] gh-143602: Fix duplicate buffer exports in io.BytesIO.write (#143629) #143872
• [3.13] gh-143602: Fix duplicate buffer exports in io.BytesIO.write (#143629) (#143872) #143878