File indexing completed on 2024-05-12 04:44:28

 #!/bin/bash
 
 # SPDX-FileCopyrightText: Lukas Sommer <sommerluk@gmail.com>
 # SPDX-License-Identifier: BSD-2-Clause OR MIT
 
 
 
 
 
 ################# Configure #################
 # The “.” command will execute the given script within the context of
 # the current script, which is necessary in order to preserve the
 # environment variables that are set by the given script.
 . scripts/export-environment.sh
 
 
 
 
 
 ################# Static code check #################
 # NOTE Keep the following list synchronized
 # between scripts/static-codecheck.sh and src/CMakeLists.txt
 PUBLIC_HEADERS="
     src/abstractdiagram.h
     src/chromahuediagram.h
     src/colordialog.h
     src/colorpatch.h
     src/colorwheel.h
     src/constpropagatinguniquepointer.h
     src/gradientslider.h
     src/importexport.h
     src/lchadouble.h
     src/lchdouble.h
     src/multispinbox.h
     src/multispinboxsection.h
     src/rgbcolorspacefactory.h
     src/settranslation.h
     src/wheelcolorpicker.h
     src/version.in.hpp
     "
 CODE_WITHOUT_UNIT_TESTS="src/* tests utils examples"
 UNIT_TESTS="autotests/*"
 ALL_CODE="$CODE_WITHOUT_UNIT_TESTS $UNIT_TESTS"
 
 EXCLUDE_PUBLIC_HEADER_FROM_GREP="--exclude=$(echo $PUBLIC_HEADERS | sed 's/ / --exclude=/g')"
 
 # TODO Test for qDebug() which should not be used never in production
 # code. Exception: Unit tests.
 
 # Search for files that do not start with a byte-order-mark (BOM).
 # We do this because Microsoft’s compiler does require a BOM at the start
 # of the file in order to interpret it as UTF-8.
 grep \
     --recursive --exclude-dir=testbed \
     --files-without-match $'\xEF\xBB\xBF' \
     $ALL_CODE \
     | grep \
         --perl-regexp "\.(license|txt|icc|qrc|qph|ts|cmake\.in|sh)$" \
         --invert-match \
     | sed 's/^/Missing byte-order-mark: /'
 
 # All header files in src/ should have an “@internal” in
 # the Doxygen documentation, because when it would be public,
 # the header file would not be in src/
 grep \
     --recursive --exclude-dir=testbed \
     $EXCLUDE_PUBLIC_HEADER_FROM_GREP \
     --files-without-match $'@internal' \
     src/*.h \
          | sed 's/^/Missing “@internal” statement in non-public header: /'
 
 # The public header files should not use “final” because it cannot be removed
 # without breaking binary compatibility.
 grep \
     --recursive --exclude-dir=testbed \
     --files-with-matches $'final' \
     $PUBLIC_HEADERS \
          | sed 's/^/“final” should not show up in public headers: /'
 
 # All public header files should use
 # the PERCEPTUALCOLOR_IMPORTEXPORT macro.
 grep \
     --recursive --exclude-dir=testbed \
     --files-without-match $'PERCEPTUALCOLOR_IMPORTEXPORT' \
     $PUBLIC_HEADERS \
          | sed 's/^/Missing PERCEPTUALCOLOR_IMPORTEXPORT macro in public header: /'
 
 # All non-public code should not use the PERCEPTUALCOLOR_IMPORTEXPORT macro.
 grep \
     --recursive --exclude-dir={testbed,include} \
     $EXCLUDE_PUBLIC_HEADER_FROM_GREP \
     --files-with-matches $'PERCEPTUALCOLOR_IMPORTEXPORT' \
     $CODE_WITHOUT_UNIT_TESTS \
          | sed 's/^/Internal files may not use PERCEPTUALCOLOR_IMPORTEXPORT macro: /'
 
 # Do not use constexpr in public headers as when we change the value
 # later, compile time value and run time value might be different, and
 # that might be dangerous.
 grep \
     --recursive --exclude-dir=testbed \
     --fixed-strings "constexpr" \
     $PUBLIC_HEADERS \
          | sed 's/^/Public headers may not use constexpr: /'
 
 # Search for some patterns that should not be used in the source code. If
 # these patterns are found, a message is displayed. Otherwise, nothing is
 # displayed.
 
 # We do not include LittleCMS headers like lcms2.h in the public API of our
 # library. But it is only be an internal dependency; library users should
 # not need to care about that. Therefore, we grab all lines that contain
 # identifiers starting with “cms” (except when in lines starting with
 # “using”). This search is not done for all code directories, but only
 # for files within the include directory (public API).
 grep \
     --recursive --exclude-dir=testbed \
     --perl-regexp "^cms" \
     $PUBLIC_HEADERS \
          | sed 's/^/Do not expose LittleCMS’ API in our public API: /'
 grep \
     --recursive --exclude-dir=testbed \
     --perl-regexp "[^a-zA-Z]cms[a-zA-Z0-9]" \
     $PUBLIC_HEADERS \
     | grep \
         --perl-regexp "\<tt\>cms" \
         --invert-match \
          | sed 's/^/Do not expose LittleCMS’ API in our public API: /'
 grep \
     --recursive --exclude-dir=testbed \
     --fixed-strings "lcms2.h" \
     $PUBLIC_HEADERS \
          | sed 's/^/Do not expose LittleCMS’ API in our public API: /'
 
 # -> Do not use the “code” and “endcode” tags for Doxygen documentation. Use
 #    @snippet instead! That allows that the example code is actually compiled
 #    and that helps detecting errors.
 grep \
     --recursive --exclude-dir=testbed \
     --perl-regexp "(@|\\\\)code(?!\{\.unparsed\})" \
     $ALL_CODE \
          | sed 's/^/Use snippets instead of “code” or “endcode”. (Exception: @code{.unparsed} is allowed.): /'
 
 # -> Doxygen style: Do not use “@em xyz”. Prefer instead “<em>xyz</em>” which
 #    might be longer, but has a clearer start point and end point, which is
 #    better when non-letter characters are involved. The @ is reserved
 #    for @ref with semantically tested references.
 # -> Same thing for “@c xyz”: Prefer instead “<tt>xyz</tt>”.
 grep \
     --recursive --exclude-dir=testbed \
     --fixed-strings "\\em" \
     $ALL_CODE \
          | sed 's/^/Use HTML tags instead of Doxygen commands for “em”: /'
 grep \
     --recursive --exclude-dir=testbed \
     --fixed-strings "@em" \
     $ALL_CODE \
          | sed 's/^/Use HTML tags instead of Doxygen commands for “em”: /'
 grep \
     --recursive --exclude-dir=testbed \
     --fixed-strings "\\c" \
     $ALL_CODE \
          | sed 's/^/Use HTML tags instead of Doxygen commands for “em”: /'
 grep \
    --recursive --exclude-dir=testbed  \
     --perl-regexp "@c(?=([^a-zA-Z]|$))" \
     $ALL_CODE \
          | sed 's/^/Use HTML tags instead of Doxygen commands for “em”: /'
 
 # -> Coding style: Do not use the “NULL” macro, but its counterpart “nullptr”
 #    which is more type save.
 grep \
     --recursive --exclude-dir=testbed \
     --fixed-strings "NULL" \
     $ALL_CODE \
          | sed 's/^/Do not use the NULL macro: /'
 
 # -> Coding style: Do not use inline functions. If used in a header,
 #    once exposed, they cannot be changed without breaking binary
 #    compatibility, because applications linking against the library
 #    will always execute the inline function version against they where
 #    compiled, and never the inline function of the library version
 #    against they are linking at run-time. This make maintaining binary
 #    compatibility much harder, for little benefit.
 #
 #    However, “static inline” variables are good for the headers or
 #    private PIMPL classes, because differently to “simple” “static”
 #    variables they allow the definition within the very same header,
 #    which is not only easier to read, but also protects against
 #    a double definition in different compiler units.
 grep \
     --recursive --exclude-dir=testbed \
     --perl-regexp "(^|[^a-zA-Z0-9\-])inline(?![^()\/;]*=.*;)" \
     $ALL_CODE \
          | sed 's/^/Do not use inline functions respectively define inline variables at the same line.: /'
 
 # -> In some Qt classes, devicePixelRatio() returns in integer.
 #    Don’t do that and use floating point precision instead. Often,
 #    devicePixelRatioF() is an alternative that provides
 #    a qreal return value.
 grep \
     --recursive --exclude-dir=testbed \
     --perl-regexp "devicePixelRatio(?!F)" \
     $CODE_WITHOUT_UNIT_TESTS \
          | sed 's/^/Use devicePixelRatioF instead of devicePixelRatio: /'
 
 # Qt’s documentation about QImage::Format says: For optimal performance only
 # use the format types QImage::Format_ARGB32_Premultiplied,
 # QImage::Format_RGB32 or QImage::Format_RGB16. Any other format, including
 # QImage::Format_ARGB32, has significantly worse performance.
 grep \
     --recursive --exclude-dir=testbed \
     --perl-regexp "QImage::Format_(?!(ARGB32_Premultiplied|RGB32|RGB16))" \
     $ALL_CODE \
          | sed 's/^/Use exclusively ARGB32_Premultiplied or RGB32 or RGB16 as QImage formats: /'
 
 # -> QLatin1String and QLatin1Char dates from the pre-Unicode epoch. It should
 #    not be used anymore at all, yet to avoid confusion. Though the Clazy
 #    codecheck suggests to use QLatin1String() instead of an empty
 #    QStringLiteral because of a slight performance gain, this might not
 #    compensate for the less readable code. Alternatively, QString() might be
 #    used to avoid Clazy warnings.
 grep \
     --recursive --exclude-dir=testbed \
     --fixed-strings "QLatin1" \
     $ALL_CODE \
          | sed 's/^/QLatin1String or QLatin1Char: If emtpy, use substitute by QString(), otherwise by QStringLiteral.: /'
 
 # When using Doxygen snippets, don’t do this within a namespace. As they are
 # meant for documentation, they should always contain fully-qualified
 # names to make sure that they always work.
 #
 # As it is complicate to know if the snippet is actually within a namespace
 # (namespaces can be opened, but also be closed again), instead we do a
 # simplified check: We simply check if there are Doxygen snippets after the
 # very first usage of the “namespace” keyword.
 #
 # 1) Display all the files that are unit tests (adding “/*” to the UNIT_TESTS
 #    variable to make sure to get a file list that “cat” will understand.
 # 2) For each file, get all lines starting from the first occurrence
 #    of “namespace”, using sed.
 # 3) Now, filter only those who actually contain snippet definitions
 #    starting with “//!”, using grep.
 for FILE in $UNIT_TESTS
 do
     cat $FILE 2>/dev/null \
         | sed -n -e '/\([^a-zA-Z]namespace\)\|\(^namespace\)/,$p' \
         | grep --fixed-strings "//!" \
         | sed 's/^/When using snippets, don’t do this after having used the “namespace” keyword: /'
 done