[gentoo-commits] repo/gentoo:master commit in: profiles/features/big-endian/, media-libs/libaom/, media-libs/libaom/files/

public inbox for gentoo-commits@lists.gentoo.org
 help / color / mirror / Atom feed

* [gentoo-commits] repo/gentoo:master commit in: profiles/features/big-endian/, media-libs/libaom/, media-libs/libaom/files/
@ 2023-10-25  3:15 Sam James
  0 siblings, 0 replies; only message in thread
From: Sam James @ 2023-10-25  3:15 UTC (permalink / raw
  To: gentoo-commits

commit:     021b730b95e79381923a3a94091f64ed98375780
Author:     Sam James <sam <AT> gentoo <DOT> org>
AuthorDate: Wed Oct 25 03:10:15 2023 +0000
Commit:     Sam James <sam <AT> gentoo <DOT> org>
CommitDate: Wed Oct 25 03:10:15 2023 +0000
URL:        https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=021b730b

media-libs/libaom: fix big-endian detection

* Backport big-endian patches
* Fix big-endian detection by passing CMake arg

The bug mentions there may still be some issues but it sounds like they're
with test data and not the tests/code itself.

Bug: https://bugs.chromium.org/p/aomedia/issues/detail?id=3487
Signed-off-by: Sam James <sam <AT> gentoo.org>

 .../files/libaom-3.7.0-big-endian-tests.patch      |  43 +++++++
 .../libaom/files/libaom-3.7.0-big-endian.patch     | 130 +++++++++++++++++++++
 .../{libaom-9999.ebuild => libaom-3.7.0-r2.ebuild} |   9 +-
 media-libs/libaom/libaom-9999.ebuild               |   6 +-
 profiles/features/big-endian/package.mask          |   4 -
 5 files changed, 186 insertions(+), 6 deletions(-)

diff --git a/media-libs/libaom/files/libaom-3.7.0-big-endian-tests.patch b/media-libs/libaom/files/libaom-3.7.0-big-endian-tests.patch
new file mode 100644
index 000000000000..38559c564726
--- /dev/null
+++ b/media-libs/libaom/files/libaom-3.7.0-big-endian-tests.patch
@@ -0,0 +1,43 @@
+https://bugs.chromium.org/p/aomedia/issues/detail?id=3487
+https://aomedia.googlesource.com/aom/+/7029529477e1473e6eb7417538cea18edc5e3bd0
+
+From 7029529477e1473e6eb7417538cea18edc5e3bd0 Mon Sep 17 00:00:00 2001
+From: Wan-Teh Chang <wtc@google.com>
+Date: Wed, 13 Sep 2023 10:55:05 -0700
+Subject: [PATCH] Fix big-endian bugs in CodingPathSync tests
+
+Change Serialize() to read uint16_t samples correctly. Although only the
+least significant byte of each sample is nonzero, we cannot assume the
+least significant byte of the c-th sample in `row` is row[c * 2]. That
+is correct only on little-endian systems.
+
+Bug: aomedia:3487
+Change-Id: I9919ce6e3c877608ca7488fe4cc6957bcfe8c4cc
+---
+ test/coding_path_sync.cc | 10 ++++++----
+ 1 file changed, 6 insertions(+), 4 deletions(-)
+
+diff --git a/test/coding_path_sync.cc b/test/coding_path_sync.cc
+index c3e51fd565..31bc2d56b0 100644
+--- a/test/coding_path_sync.cc
++++ b/test/coding_path_sync.cc
+@@ -130,11 +130,13 @@ std::vector<int16_t> Serialize(const aom_image_t *img) {
+ 
+     for (int r = 0; r < h; ++r) {
+       for (int c = 0; c < w; ++c) {
+-        unsigned char *row = img->planes[plane] + r * img->stride[plane];
+-        if (img->fmt & AOM_IMG_FMT_HIGHBITDEPTH)
+-          bytes.push_back(row[c * 2]);
+-        else
++        const unsigned char *row = img->planes[plane] + r * img->stride[plane];
++        if (img->fmt & AOM_IMG_FMT_HIGHBITDEPTH) {
++          const uint16_t *row16 = reinterpret_cast<const uint16_t *>(row);
++          bytes.push_back(row16[c]);
++        } else {
+           bytes.push_back(row[c]);
++        }
+       }
+     }
+   }
+-- 
+2.42.0

diff --git a/media-libs/libaom/files/libaom-3.7.0-big-endian.patch b/media-libs/libaom/files/libaom-3.7.0-big-endian.patch
new file mode 100644
index 000000000000..907e5fd1110e
--- /dev/null
+++ b/media-libs/libaom/files/libaom-3.7.0-big-endian.patch
@@ -0,0 +1,130 @@
+https://bugs.chromium.org/p/aomedia/issues/detail?id=3487
+https://aomedia.googlesource.com/aom/+/8b65d76f2f95b1bfdc82b8f135246cb367d444e6%5E%21/
+
+From 8b65d76f2f95b1bfdc82b8f135246cb367d444e6 Mon Sep 17 00:00:00 2001
+From: Wan-Teh Chang <wtc@google.com>
+Date: Tue, 12 Sep 2023 16:38:47 -0700
+Subject: [PATCH] Fix byte order issue in write_enc_data_to_out_buf
+
+Fix the byte order issue for big-endian systems introduced in
+https://aomedia-review.googlesource.com/c/aom/+/167581.
+
+Use the existing HToBE64 macro defined in aom_util/endian_inl.h and
+delete the new get_byteswap64 function from aom_ports/bitops.h.
+
+Based on the patch in bug aomedia:3487.
+
+Bug: aomedia:3487
+Change-Id: I56183e062a9f2b18c4c6158fd8e47c5062de85ee
+---
+ aom_dsp/entenc.h   |  7 ++++---
+ aom_ports/bitops.h | 35 +----------------------------------
+ 2 files changed, 5 insertions(+), 37 deletions(-)
+
+diff --git a/aom_dsp/entenc.h b/aom_dsp/entenc.h
+index 467e47bf56..d26f027ed0 100644
+--- a/aom_dsp/entenc.h
++++ b/aom_dsp/entenc.h
+@@ -13,7 +13,7 @@
+ #define AOM_AOM_DSP_ENTENC_H_
+ #include <stddef.h>
+ #include "aom_dsp/entcode.h"
+-#include "aom_ports/bitops.h"
++#include "aom_util/endian_inl.h"
+ 
+ #ifdef __cplusplus
+ extern "C" {
+@@ -87,13 +87,14 @@ static AOM_INLINE void propagate_carry_bwd(unsigned char *buf, uint32_t offs) {
+   } while (carry);
+ }
+ 
+-// Reverse byte order and write data to buffer adding the carry-bit
++// Convert to big-endian byte order and write data to buffer adding the
++// carry-bit
+ static AOM_INLINE void write_enc_data_to_out_buf(unsigned char *out,
+                                                  uint32_t offs, uint64_t output,
+                                                  uint64_t carry,
+                                                  uint32_t *enc_offs,
+                                                  uint8_t num_bytes_ready) {
+-  const uint64_t reg = get_byteswap64(output) >> ((8 - num_bytes_ready) << 3);
++  const uint64_t reg = HToBE64(output << ((8 - num_bytes_ready) << 3));
+   memcpy(&out[offs], &reg, 8);
+   // Propagate carry backwards if exists
+   if (carry) {
+diff --git a/aom_ports/bitops.h b/aom_ports/bitops.h
+index 3c5b992bde..7f4c165f5a 100644
+--- a/aom_ports/bitops.h
++++ b/aom_ports/bitops.h
+@@ -13,7 +13,6 @@
+ #define AOM_AOM_PORTS_BITOPS_H_
+ 
+ #include <assert.h>
+-#include <stdint.h>
+ 
+ #include "aom_ports/msvc.h"
+ #include "config/aom_config.h"
+@@ -34,12 +33,8 @@ extern "C" {
+ // These versions of get_msb() are only valid when n != 0 because all
+ // of the optimized versions are undefined when n == 0:
+ 
+-// get_byteswap64:
+-// Returns the number (uint64_t) with byte-positions reversed
+-// e.g. input 0x123456789ABCDEF0 returns 0xF0DEBC9A78563412
+-
+ // GCC compiler: https://gcc.gnu.org/onlinedocs/gcc/Other-Builtins.html
+-// MSVC: https://learn.microsoft.com/en-us/cpp/c-runtime-library/
++// MSVC: https://learn.microsoft.com/en-us/cpp/intrinsics/compiler-intrinsics
+ 
+ // use GNU builtins where available.
+ #if defined(__GNUC__) && \
+@@ -48,10 +43,6 @@ static INLINE int get_msb(unsigned int n) {
+   assert(n != 0);
+   return 31 ^ __builtin_clz(n);
+ }
+-
+-static INLINE uint64_t get_byteswap64(uint64_t num) {
+-  return __builtin_bswap64(num);
+-}
+ #elif defined(USE_MSC_INTRINSICS)
+ #pragma intrinsic(_BitScanReverse)
+ 
+@@ -61,10 +52,6 @@ static INLINE int get_msb(unsigned int n) {
+   _BitScanReverse(&first_set_bit, n);
+   return first_set_bit;
+ }
+-
+-static INLINE uint64_t get_byteswap64(uint64_t num) {
+-  return _byteswap_uint64(num);
+-}
+ #undef USE_MSC_INTRINSICS
+ #else
+ static INLINE int get_msb(unsigned int n) {
+@@ -82,26 +69,6 @@ static INLINE int get_msb(unsigned int n) {
+   }
+   return log;
+ }
+-
+-static INLINE uint64_t get_byteswap64(uint64_t num) {
+-  uint64_t out = 0x00;
+-  uint64_t mask = 0xFF00000000000000;
+-  int bit_shift = 56;  // 7 bytes
+-  // 4 ms bytes
+-  do {
+-    out |= (num & mask) >> bit_shift;
+-    mask >>= 8;
+-    bit_shift -= 16;
+-  } while (bit_shift >= 0);
+-  // 4 ls bytes
+-  bit_shift = 8;  // 1 byte
+-  do {
+-    out |= (num & mask) << bit_shift;
+-    mask >>= 8;
+-    bit_shift += 16;
+-  } while (bit_shift <= 56);
+-  return out;
+-}
+ #endif
+ 
+ #ifdef __cplusplus
+-- 
+2.42.0

diff --git a/media-libs/libaom/libaom-9999.ebuild b/media-libs/libaom/libaom-3.7.0-r2.ebuild
similarity index 92%
copy from media-libs/libaom/libaom-9999.ebuild
copy to media-libs/libaom/libaom-3.7.0-r2.ebuild
index 76a2b4d37365..3fbef9beb31c 100644
--- a/media-libs/libaom/libaom-9999.ebuild
+++ b/media-libs/libaom/libaom-3.7.0-r2.ebuild
@@ -33,7 +33,7 @@ HOMEPAGE="https://aomedia.org https://aomedia.googlesource.com/aom/"
 
 LICENSE="BSD-2"
 SLOT="0/3"
-IUSE="doc +examples test"
+IUSE="big-endian doc +examples test"
 IUSE="${IUSE} cpu_flags_x86_mmx cpu_flags_x86_sse cpu_flags_x86_sse2 cpu_flags_x86_sse3 cpu_flags_x86_ssse3"
 IUSE="${IUSE} cpu_flags_x86_sse4_1 cpu_flags_x86_sse4_2 cpu_flags_x86_avx cpu_flags_x86_avx2"
 IUSE="${IUSE} cpu_flags_arm_neon"
@@ -57,6 +57,10 @@ DOCS=( PATENTS )
 
 PATCHES=(
 	"${FILESDIR}"/${PN}-3.4.0-posix-c-source-ftello.patch
+	"${FILESDIR}"/${PN}-3.7.0-allow-fortify-source.patch
+
+	"${FILESDIR}"/${P}-big-endian.patch
+	"${FILESDIR}"/${P}-big-endian-tests.patch
 )
 
 multilib_src_configure() {
@@ -68,6 +72,9 @@ multilib_src_configure() {
 		-DENABLE_TESTS=$(usex test)
 		-DENABLE_TOOLS=ON
 		-DENABLE_WERROR=OFF
+		# https://bugs.chromium.org/p/aomedia/issues/detail?id=3487 shows
+		# that big endian detection doesn't really work otherwise.
+		-DCONFIG_BIG_ENDIAN=$(usex big-endian 1 0)
 
 		# Needs libjxl, currently unpackaged.
 		-DCONFIG_TUNE_BUTTERAUGLI=0

diff --git a/media-libs/libaom/libaom-9999.ebuild b/media-libs/libaom/libaom-9999.ebuild
index 76a2b4d37365..80b545dbde32 100644
--- a/media-libs/libaom/libaom-9999.ebuild
+++ b/media-libs/libaom/libaom-9999.ebuild
@@ -33,7 +33,7 @@ HOMEPAGE="https://aomedia.org https://aomedia.googlesource.com/aom/"
 
 LICENSE="BSD-2"
 SLOT="0/3"
-IUSE="doc +examples test"
+IUSE="big-endian doc +examples test"
 IUSE="${IUSE} cpu_flags_x86_mmx cpu_flags_x86_sse cpu_flags_x86_sse2 cpu_flags_x86_sse3 cpu_flags_x86_ssse3"
 IUSE="${IUSE} cpu_flags_x86_sse4_1 cpu_flags_x86_sse4_2 cpu_flags_x86_avx cpu_flags_x86_avx2"
 IUSE="${IUSE} cpu_flags_arm_neon"
@@ -57,6 +57,7 @@ DOCS=( PATENTS )
 
 PATCHES=(
 	"${FILESDIR}"/${PN}-3.4.0-posix-c-source-ftello.patch
+	"${FILESDIR}"/${PN}-3.7.0-allow-fortify-source.patch
 )
 
 multilib_src_configure() {
@@ -68,6 +69,9 @@ multilib_src_configure() {
 		-DENABLE_TESTS=$(usex test)
 		-DENABLE_TOOLS=ON
 		-DENABLE_WERROR=OFF
+		# https://bugs.chromium.org/p/aomedia/issues/detail?id=3487 shows
+		# that big endian detection doesn't really work otherwise.
+		-DCONFIG_BIG_ENDIAN=$(usex big-endian 1 0)
 
 		# Needs libjxl, currently unpackaged.
 		-DCONFIG_TUNE_BUTTERAUGLI=0

diff --git a/profiles/features/big-endian/package.mask b/profiles/features/big-endian/package.mask
index e85ed1dfbe00..f13c23ad9e48 100644
--- a/profiles/features/big-endian/package.mask
+++ b/profiles/features/big-endian/package.mask
@@ -11,10 +11,6 @@ games-engines/fheroes2
 games-engines/openmw
 games-strategy/homm2-gold-gog
 
-# Sam James <sam@gentoo.org> (2023-09-28)
-# Broken on BE (https://bugs.chromium.org/p/aomedia/issues/detail?id=3487)
-~media-libs/libaom-3.7.0
-
 # matoro <matoro_gentoo@matoro.tk> (2023-06-30)
 # Test failures on BE
 # https://github.com/libjxl/libjxl/issues/2433


^ permalink raw reply related	[flat|nested] only message in thread

only message in thread, other threads:[~2023-10-25  3:15 UTC | newest]

Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2023-10-25  3:15 [gentoo-commits] repo/gentoo:master commit in: profiles/features/big-endian/, media-libs/libaom/, media-libs/libaom/files/ Sam James

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox