Debugging build crashes in some circumstances in str.decode() with error handler which produces replacement string with length larger than malformed data. For example the backslashreplace error handler produces 4-character string for every illegal byte. All other standard error handlers produce no longer than 1 character for every illegal unit.

Here is a patch which fixes this issue. I'll commit it without review because buildbots are broken without it. This issue is open for reference and post-commit review.

New changeset 2de90090e486 by Serhiy Storchaka in branch '3.4':
Issue #23321: Fixed a crash in str.decode() when error handler returned
https://hg.python.org/cpython/rev/2de90090e486

New changeset 1cd68b3c46aa by Serhiy Storchaka in branch 'default':
Issue #23321: Fixed a crash in str.decode() when error handler returned
https://hg.python.org/cpython/rev/1cd68b3c46aa

> Debugging build crashes in some circumstances in str.decode() (...) buildbots are broken without it

Is it a regression? Would it be possible to identify the changeset
responsible of the regression?

I think the changeset which made decoders to use _PyUnicodeWriter (issue16311) 
is responsible of the regression.

For example consider b'\x80abc'.decode('utf-8', 'backslashreplace').

The writer reserves string buffer with size 4 (every byte produces at most 1 
character). First byte is incorrect and replaced by 4-character string 
'\\x80'. The writer increases min_length but doesn't resize the buffer because 
its size is enough to write replacement string. But following writes of ASCII 
characters cause buffer overflow.

New changeset 1e8937861ee3 by Victor Stinner in branch 'default':
Issue #22286, #23321: Fix failing test on Windows code page 932
https://hg.python.org/cpython/rev/1e8937861ee3

If you have no enhancements to my quick fix Victor, may be this issue can be closed.

I closed the issue.