Second argument of Tcl_NewUnicodeObj() which specifies a number of characters has type "int". When a large string with more than INT_MAX characters passed to Tkinter this value will overflow. If this parameter is negative, all characters up to the first null character are used (up to the end of string because current string implementation has null character after the end). When string length will be more than UINT_MAX, always only part of string will be converted.

I propose explicitly check a string length and raise an exception when the length overflows C int range.

Here is a patch. Please run test_tcl test on 64-bit platform with > 20 GiB of memory.

I haven't included a test for huge tuples because it requires a lot of memory (perhaps the hundreds of GiBs) and should run a long time. It is impossible to calculate exact requirements from Python side, it's TCL implementation details. I'm not sure even about a huge string test.

New changeset 6bc533d06cf1 by Serhiy Storchaka in branch '3.3':
Issue #17119: Fixed integer overflows when processing large strings and tuples
http://hg.python.org/cpython/rev/6bc533d06cf1

New changeset b500daaee7d0 by Serhiy Storchaka in branch 'default':
Issue #17119: Fixed integer overflows when processing large strings and tuples
http://hg.python.org/cpython/rev/b500daaee7d0

New changeset ee0bb673536c by Serhiy Storchaka in branch '2.7':
Issue #17119: Fixed integer overflows when processing large Unicode strings
http://hg.python.org/cpython/rev/ee0bb673536c