requires them. Disable executable bits and shebang lines in test and
benchmark files in order to prevent using a random system python, and in
source files of modules which don't provide command line interface. Fixed
shebang line to use python3 executable in the unittestgui script.
This fixes an edge case (20206) where if the input ended in a character
needing encoding but there was no newline on the string, the last byte
of the encoded character would be dropped. The fix is to use a more
efficient algorithm, provided by Serhiy Storchaka (5803), that does not
have the bug.
* The first line of Python script could be executed twice when the source
encoding (not equal to 'utf-8') was specified on the second line.
* Now the source encoding declaration on the second line isn't effective if
the first line contains anything except a comment.
* As a consequence, 'python -x' works now again with files with the source
encoding declarations specified on the second file, and can be used again
to make Python batch files on Windows.
* The tokenize module now ignore the source encoding declaration on the second
line if the first line contains anything except a comment.
* IDLE now ignores the source encoding declaration on the second line if the
first line contains anything except a comment.
* 2to3 and the findnocoding.py script now ignore the source encoding
declaration on the second line if the first line contains anything except
a comment.
is modified during the lifetime of the Python process after zipimport has
already cached the zip's table of contents we detect this and recover
rather than read bad data from the .zip (causing odd import errors).
In issue 20074 it was pointed out that getpass would fail with a traceback if
stdin was, for example /dev/null, which is a non-unlikely scenario.
Also backported the tests from issue 17484 as modified by issue 18116.
(What I really did was copy getpass.py and test_getpass.py from their
state on tip as of 17bd04fbf3d3).
acceptable outcomes: success or RuntimeError. Without being able to
actually hear whether a sound was played, either one could be right, but
any other error would be a failure.
* splitunc() no more return illegal result for paths with redundant slashes.
* splitunc() now correctly processes the 'İ' character
(U+0130, LATIN CAPITAL LETTER I WITH DOT ABOVE).
* Deprecation warnings now emitted for every use of splitunc().
* Added tests for splitunc().
