Change possible_levels to a tuple. Arrays are mutable,
but we never want to mutate this value, so changing it
to a tuple reduces the risk of accidentally mutating it.
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
This commit means that the output provides useful summary data,
even if the lower-level hits are suppressed.
Note that this does use a little more memory when some hits
are supressed, since the hitlist is fully created even
if only parts are displayed. However, modern systems have
lots of memory. Hopefully we'll never analyze software
with so many problems that this is a problem itself :-).
If someone ever has that problem, they can output everything
and filter it separately.
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
My upload intended for pypitest appears to have gone to pypi instead.
To eliminate confusion, I'm bumping the version number so that
any single version number always refers to exactly one program version.
This was done with:
sed -i.bak -e 's/2\.0\.3/2.0.4/g' \
ChangeLog correct-results.* flawfinder makefile setup.py
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
Test the saving and loading of hitlists.
This detected a Python3 problem, which was easily corrected by
saving and loading in binary "b" format instead of text format.
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
We formerly used a variable named "input". This is legal in Python,
but potentially confusing since there's a built-in named "input"
that this shadows in that scope. Rename the variable, to
avoid that confusion. This fixes the following pylint warning:
W:1440, 8: Redefining built-in 'input' (redefined-builtin)
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
We import os, so there's no need to import os.path.
This fixes the following pylint warning:
C: 53, 0: Imports from package os are not grouped (ungrouped-imports)
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
Python 3 only accepts certain syntaxes for find & split.
Thankfully, it's possible to use them in Python 2, so rework
it so we can use the same syntax for both.
This is not detected or fixed by futurize, sadly
(a problem true for many other situations).
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
Remove use of __cmp__, which is in Python 2 but not in Python 3.
Instead, use sort keys, which work in Python 2 and 3.
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
Switch all print statements to print() functions per PEP 3105.
Python 3 *only* supports print() functions, so this begins to
move the code towards simultaneously supporting python 2 and 3.
This implements "stage1" of futurize. In theory, "stage1" is
supposed to be "low risk", but in fact a *large* number of
manual fixes had to be made to make the program work again.
Python 2's traditional print statement includes the "softspace"
feature. This is "a semi-secret attribute on files currently used to tell
print whether to insert a space before the first item". The print()
function does not have the "softspace" feature, so there is no direct
translation for any situation that depended on softspaces.
Flawfinder used softspaces extensively, as they were convenient,
so it took a little work to make print() functions work.
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
Pylint is a static analyzer of Python code. It seems appropriate
to use a static analyzer to analyze a static analyzer :-).
None of the changes here fix a flaw in flawfinder.
The goal here is to follow PEP 008 more closely in the hope that
the changes make it easier for other developers to improve it further.
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
This switches flawfinder's version naming conventions to comply with
Semantic Versioning. We change the first digit to "2" because
there's a subtle change in how CWEs are reported - see the
ChangeLog or documentation for more information.
Signed-off-by: David A. Wheeler <dwheeler@dwheeler.com>
- snprintf is a useful *countermeasure* for buffer overflows,
and unlike some alternatives it is standard and *widely* available.
(strlcpy/strlcat are useful but not standard and not widely available;
snprintf_s is standard but not widely available).
Historically we warned about snprintf because old systems didn't
implement it correctly, but at this point these old systems are
more historical than anything else.
Instead, let's specifically *mention* snprintf as a recommended
potential solution for buffer overflows.
- Provide a separate warning for filenames listed on the command line
but do not actually exist.
- Provide a SPECIAL warning for filenames that begin with
the UTF-8 sequence of long dashes. Users might
copy-and-paste from the pdf version of the man page, and that might
produce long dashes (instead of normal dashes). This problem
can be hard to detect, so specially warning about it should help.
David A. Wheeler
Jon Hood
David A. Wheeler