Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
---|---|---|---|---|---|---|
0001-Accept-Debian-build-flags.patch | Accept Debian build flags | YOKOTA Hiroshi <yokota.hgml@gmail.com> | not-needed | 2021-05-09 | ||
0002-Disable-hardware-acceleration-support-on-armel.patch | Disable hardware acceleration support on armel Use "__ARM_ARCH" to split "armel" and "armhf". ARM v7 feature breaks ARM v5 compiler because the feature is not supported by ARM v5 compiler. __ARM_ARCH: armel = 5 armhf = 7 |
YOKOTA Hiroshi <yokota.hgml@gmail.com> | yes | 2021-09-14 | ||
0003-Guard-ARM-v8-feature-from-old-architecture.patch | Guard ARM v8 feature from old architecture | YOKOTA Hiroshi <yokota.hgml@gmail.com> | yes | 2021-10-13 | ||
0004-Use-getcwd-3-POSIX-extension-to-avoid-PATH_MAX-macro.patch | Use getcwd(3) POSIX extension to avoid PATH_MAX macro This fix helps GNU Hurd. |
YOKOTA Hiroshi <yokota.hgml@gmail.com> | yes | 2021-09-15 | ||
0005-Disable-local-echo-display-when-in-input-passwords-C.patch | Disable local echo display when in input passwords (Closes: #1006238) | YOKOTA Hiroshi <yokota.hgml@gmail.com> | yes | debian | 2022-02-22 | |
0006-Fix-ARM-architecture-selection-code.patch | Fix ARM architecture selection code This fix guards NEON function usage from older ARM architectures like armel (__ARM_ARCH == 5). |
YOKOTA Hiroshi <yokota.hgml@gmail.com> | yes | 2023-06-21 | ||
0007-Add-__target__-fpu-neon-attribute-to-use-NEON-specif.patch | Add __target__("fpu=neon") attribute to use NEON specific codes https://sourceforge.net/p/sevenzip/patches/413/#8bce > Igor Pavlov: > 7-zip detects neon availability at runtime via hwcap. > And 7-zip can select between neon and non-neon code. > So we want to use NEON in some functions, even if main code is not neon. > for x86, we just use `__attribute__` for avx2 functions: > > ``` > __attribute__((__target__("avx2"))) > ``` > > What way to do similar thing for arm-neon ? From armhf build log: > /usr/lib/gcc/arm-linux-gnueabihf/12/include/arm_neon.h:9186:1: > error: inlining failed in call to ‘always_inline’ ‘vrev16q_u8’: > target specific option mismatch vrev16q_u8() requires NEON option to use. Add __target__("fpu=neon") attribute to enable NEON intrinsics. __target__("fpu=crypto-neon-fp-armv8") is too much to use vrev16q_u8(). Use simple __target__("fpu=neon") to satisfy lowest requirement to use vrev16q_u8(). |
YOKOTA Hiroshi <yokota.hgml@gmail.com> | yes | 2023-06-21 | ||
0008-PPC-is-already-defined-as-1-on-PowerPC-compilers.patch | "PPC" is already defined as "1" on PowerPC compilers Use "Z7_PPC" to avoid symbol name conflicts |
YOKOTA Hiroshi <yokota.hgml@gmail.com> | yes | 2023-06-21 | ||
0009-64bit-time-data-type-fix.patch | 64bit time data type fix Use C++17 function timespec_get() to get current UTC time. On 32bit architecture (armel,armhf,m68k,powerpc,sh4): ``` ../../../Windows/TimeUtils.cpp: In function ‘void NWindows::NTime::GetCurUtc_FiTime(timespec&)’: ../../../Windows/TimeUtils.cpp:264:30: error: conversion from ‘__suseconds64_t’ {aka ‘long long int’} to ‘long int’ may change value [-Werror=conversion] 264 | ft.tv_nsec = now.tv_usec * 1000; | ~~~~~~~~~~~~^~~~~~ ``` |
YOKOTA Hiroshi <yokota.hgml@gmail.com> | yes | 2024-04-11 |