For the former case above, the line doesn't include a newline which
means we never parse a 'NEWLINE' token with a logical line (the fifth
element of the token tuple) corresponding to our full line. This means
we don't here but that wasn't an issue previously since there's a
fallthrough case that handled tokens remaining at the end of the parse:
Unfortunately, because we now have an additional newline character to
parse, one that's on a separate line to our test string no less, we run
logical checks on it:
This fixes changes things so that a newline is included (and also adds
quotes so it's valid Python, but that's mostly unrelated). This means we
end up with the following instead:
["'This is the the best comment'\n"]
On both Python without the bugfix and with it, this parses as:
(3, "'This is the the best comment'", (1, 0), (1, 30), "'This is the the best comment'\n")
(4, '\n', (1, 30), (1, 31), "'This is the the best comment'\n")
(0, '', (2, 0), (2, 0), '')
where (3 = 'STRING', 4 = 'NEWLINE', 0 = 'ENDMARKER')
This isn't _really_ a fix since there's clearly still a bug in either
'pycodestyle' or Python (I think the latter, since it's adding a newline
to a file that explicitly doesn't have one), but the chances of us
hitting this bug in practice are rather low - you'd need to make a
mistake on the very last line of a file without a newline at the end
which is something Vim, for example, won't even let you do without
setting special flags - and therefore it can be reasonably ignored.
Reviewed: https:/ /review. opendev. org/664940 /git.openstack. org/cgit/ openstack/ nova/commit/ ?id=f545a25cc44 3c41dcd9bdd0280 64c28b53f56037
Committed: https:/
Submitter: Zuul
Branch: master
commit f545a25cc443c41 dcd9bdd028064c2 8b53f56037
Author: Stephen Finucane <email address hidden>
Date: Wed Jun 12 15:10:59 2019 +0100
Fix double word hacking test
At present, 'pycodestyle' feeds the following string into the 'tokenizer'
library:
["'This is the the best comment'"]
(note the added quotes because this isn't valid Python otherwise)
On previous versions of Python, this tokenizer would parse the string like so:
(3, "'This is the the best comment'", (1, 0), (1, 30), "'This is the the best comment'")
(0, '', (2, 0), (2, 0), '')
where (3 = 'STRING', 0 = 'ENDMARKER')
However, with the fix [1] backported to recent versions of Python, this now
resolves to:
(3, "'This is the the best comment'", (1, 0), (1, 30), "'This is the the best comment'")
(4, '', (1, 30), (1, 31), '')
(0, '', (2, 0), (2, 0), '')
where (3 = 'STRING', 4 = 'NEWLINE', 0 = 'ENDMARKER')
Typically, 'pycodestyle' will run physical line checks on each line as it
parses the token:
https:/ /github. com/PyCQA/ pycodestyle/ blob/2. 5.0/pycodestyle .py#L2036
For the former case above, the line doesn't include a newline which
means we never parse a 'NEWLINE' token with a logical line (the fifth
element of the token tuple) corresponding to our full line. This means
we don't here but that wasn't an issue previously since there's a
fallthrough case that handled tokens remaining at the end of the parse:
https:/ /github. com/PyCQA/ pycodestyle/ blob/2. 5.0/pycodestyle .py#L2114- L2116
Unfortunately, because we now have an additional newline character to
parse, one that's on a separate line to our test string no less, we run
logical checks on it:
https:/ /github. com/PyCQA/ pycodestyle/ blob/2. 5.0/pycodestyle .py#L2105- L2107
This is an issue since the logical check wipes stored tokens meaning
we've nothing to check when we get to the fallthrough case:
https:/ /github. com/PyCQA/ pycodestyle/ blob/2. 5.0/pycodestyle .py#L2012
This fixes changes things so that a newline is included (and also adds
quotes so it's valid Python, but that's mostly unrelated). This means we
end up with the following instead:
["'This is the the best comment'\n"]
On both Python without the bugfix and with it, this parses as:
(3, "'This is the the best comment'", (1, 0), (1, 30), "'This is the the best comment'\n")
(4, '\n', (1, 30), (1, 31), "'This is the the best comment'\n")
(0, '', (2, 0), (2, 0), '')
where (3 = 'STRING', 4 = 'NEWLINE', 0 = 'ENDMARKER')
Which triggers things in 'pycodestyle' correctly.
https:/ /github. com/PyCQA/ pycodestyle/ blob/2. 5.0/pycodestyle .py#L2044- L2046
This isn't _really_ a fix since there's clearly still a bug in either
'pycodestyle' or Python (I think the latter, since it's adding a newline
to a file that explicitly doesn't have one), but the chances of us
hitting this bug in practice are rather low - you'd need to make a
mistake on the very last line of a file without a newline at the end
which is something Vim, for example, won't even let you do without
setting special flags - and therefore it can be reasonably ignored.
[1] https:/ /bugs.python. org/issue33899
Change-Id: Ia597594e0469c0 e83d7ad22b06783 90aaebaffe7
Signed-off-by: Stephen Finucane <email address hidden>
Closes-Bug: #1804062