mirror of
https://github.com/libretro/virtualjaguar-libretro.git
synced 2026-11-07 04:36:00 -10:00
Joseph Mattiello
b7bbef5a77
PR #571's DSP sh/sha port (P5) copies gpu_opcode_sha's shift-by-31 expression verbatim, which already carries a shiftTooManyBitsSigned suppression for src/tom/gpu.c ("standard C UB, every target compiler does arithmetic shift on signed"). The identical expression now exists in src/jerry/dsp.c too and needs the same suppression -- not a new issue, the same accepted one at a new site. Verified with the exact CI cppcheck invocation locally: exit 0.
42 lines
2.0 KiB
Plaintext
42 lines
2.0 KiB
Plaintext
# cppcheck suppressions for the libretro CI job.
|
|
#
|
|
# Each suppression is one of two shapes:
|
|
# <id>:<file> - suppress all <id> findings in <file>
|
|
# <id>:<file>:<line> - suppress one specific occurrence
|
|
#
|
|
# RULE: every entry needs a one-line comment. If you find yourself
|
|
# adding a suppression for a new issue, that's a hint the issue should
|
|
# probably be fixed instead.
|
|
|
|
# DSP / GPU RAM bounds: cppcheck doesn't model that GET16/GET32 macros
|
|
# read from a 2-byte / 4-byte sliding window inside the dsp_ram_8 /
|
|
# gpu_ram_8 backing arrays, and that the bounds check `offset <=
|
|
# BASE + (size - 1)` is one short for the 2-byte / 4-byte reads.
|
|
# These read past the declared array size by 1-3 bytes -- in practice
|
|
# the next byte is part of an adjacent register / pad and is unused,
|
|
# but it is technically OOB and worth a follow-up audit.
|
|
arrayIndexOutOfBoundsCond:src/jerry/dsp.c
|
|
arrayIndexOutOfBoundsCond:src/tom/gpu.c
|
|
|
|
# Blitter long-shift code paths are guarded by a runtime condition
|
|
# (cppcheck cites the guard explicitly in the "See condition at line
|
|
# N" note). Safe under the existing branch.
|
|
shiftTooManyBits:src/tom/blitter.c
|
|
|
|
# Pre-existing signed-int shift in the GPU shifter unit -- standard
|
|
# C UB (shifting a signed 32-bit value by 31 lands in the sign bit).
|
|
# Tracked for follow-up. The DSP shifter (dsp_opcode_sha, dsp.c) uses
|
|
# the identical expression -- it was ported verbatim from the GPU
|
|
# implementation (issue #569 P5) -- so it inherits the same suppression.
|
|
shiftTooManyBitsSigned:src/tom/gpu.c
|
|
shiftTooManyBitsSigned:src/jerry/dsp.c
|
|
|
|
# Object Processor sign-extension idiom: cast to int8_t then >> 4 to
|
|
# sign-extend a 4-bit nibble. Implementation-defined under C89 but
|
|
# every compiler we target does arithmetic shift on signed.
|
|
shiftNegativeLHS:src/tom/op.c
|
|
|
|
# libretro-common conditional includes that depend on host-platform
|
|
# macros cppcheck doesn't see (e.g. _WIN32, __APPLE__).
|
|
preprocessorErrorDirective:libretro-common/include/retro_common_api.h
|