Upgrade GCC 13.2.0 version (PPU-SPU) by humbertodias · Pull Request #121 · ps3dev/ps3toolchain

humbertodias · 2024-03-05T01:16:34Z

Features

binutils-ppu bumped from 2.22 to 2.42
ppu-gcc bumped from 7.20 to 13.2.0
spu-gcc bumped from 7.20 to 9.5.0
docker bumped from 16.04 to 24.04 - LTS
ci/cd working again and building releases
ci/cd to docker hub
ci/cd badges
build on ubuntu/macOS for arm64 enabled
test PSL1GHT hello game

Note

We've selected the version 9.5.0 because it's the last GCC release to include support for SPU.
https://www.phoronix.com/news/GCC-10-Drops-Cell-BE-SPU

Tip

Patches credits to disc-kuraudo and luizfernandonb

Result

ar --version

GNU ar (GNU Binutils) 2.42
Copyright (C) 2024 Free Software Foundation, Inc.
This program is free software; you may redistribute it under the terms of
the GNU General Public License version 3 or (at your option) any later version.
This program has absolutely no warranty.

ppu-gcc --version

ppu-gcc (GCC) 13.2.0
Copyright (C) 2023 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

spu-gcc --version

spu-gcc (GCC) 9.5.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Docker

FROM ubuntu:24.04

ci/cd success

https://github.com/humbertodias/ps3toolchain/actions/runs/8166663270

ci/cd docker

Note

Don't forget to create a named DockerHub environment on your repository with two secret variables: DOCKERHUB_USERNAME and DOCKERHUB_TOKEN

badges
PSL1GHT hello game

Fewnity · 2024-03-05T18:00:54Z

Hey cool work, I really want gcc 13 on ps3. But you just forgot to credit Darjan Krijanand and luizfernandonb :/

humbertodias · 2024-03-05T18:58:56Z

Hey cool work, I really want gcc 13 on ps3. But you just forgot to credit Darjan Krijanand and luizfernandonb :/

Yeah.. I left his name on patch's file given him the credit

void2012 · 2026-09-01T15:35:33Z

Not like I care about PS3 anymore, but I appreciate your effort. It's insane that we could use C++20 for PS3 now. Thanks anyway.

TheMrIron2 · 2026-09-01T15:49:59Z

Thanks for your work @humbertodias . Several of us maintainers greatly appreciate the work you've done for upgrading gcc on PS3 and for coming back to update us. I hope void's reply did not dissuade you.

Do you intend to do any further work on this branch? Or should we consider this ready for review?

zeldin · 2026-09-01T16:37:49Z

A question: Which commits are the patches generated from? I don't see corresponding pull requests to https://github.com/ps3dev/gcc-PS3 et al. If new base repos are used (which might make sense given a large jump in version), it would be good if those could be moved to the ps3dev org for maintainability.

Cruslan · 2026-09-02T01:58:16Z

When building on a modern Linux distribution with recent GCC 16, the toolchain build fails at step 002-gcc-newlib-PPU.sh with two distinct issues:

1. `libcody` UTF-8 / char8_t type mismatch

GCC 16 enables -fchar8_t by default in C++20 mode, causing u8"" string literals in libcody/client.cc to fail implicit conversions against std::string:

../../libcody/client.cc:329:42: error: invalid conversion from 'const char8_t*' to 'size_t' {aka 'long unsigned int'} [-fpermissive]
  329 |         return Packet (Client::PC_ERROR, u8"");

2. `safe-ctype.h` macro poisoning in `gcc/system.h`

The current gcc-13.2.0-PS3-macos.patch is restricted strictly to Darwin (if [[ $(uname -s) == 'Darwin' ]]). However, modern libstdc++ on Linux also pulls <locale> and <memory> through standard headers, resulting in islower / toupper macro conflicts during the compilation of c-decl.cc:

In file included from /usr/include/g++-v16/bits/locale_facets.h:2639,
                 from ../../gcc/c/c-decl.cc:
/usr/include/g++-v16/bits/locale_facets.h:2639:5: error: 'template<class _CharT> bool std::islower' conflicts with a previous declaration
 2639 |     islower(_CharT __c, const locale& __loc)

Moreover, guarding headers behind #ifdef INCLUDE_STRING inside system.h is insufficient because several C front-end source files pull <locale> transitively without explicitly defining INCLUDE_STRING.

It seems fixed for macOS builds but the same patch not applied for Linux hosts. Before than merge it has to be fixed.

humbertodias · 2026-09-02T14:21:11Z

When building on a modern Linux distribution with recent GCC 16, the toolchain build fails at step 002-gcc-newlib-PPU.sh with two distinct issues:

1. libcody UTF-8 / char8_t type mismatch

GCC 16 enables -fchar8_t by default in C++20 mode, causing u8"" string literals in libcody/client.cc to fail implicit conversions against std::string:
../../libcody/client.cc:329:42: error: invalid conversion from 'const char8_t*' to 'size_t' {aka 'long unsigned int'} [-fpermissive]
  329 |         return Packet (Client::PC_ERROR, u8"");
2. safe-ctype.h macro poisoning in gcc/system.h

The current gcc-13.2.0-PS3-macos.patch is restricted strictly to Darwin (if [[ $(uname -s) == 'Darwin' ]]). However, modern libstdc++ on Linux also pulls <locale> and <memory> through standard headers, resulting in islower / toupper macro conflicts during the compilation of c-decl.cc:
In file included from /usr/include/g++-v16/bits/locale_facets.h:2639,
                 from ../../gcc/c/c-decl.cc:
/usr/include/g++-v16/bits/locale_facets.h:2639:5: error: 'template<class _CharT> bool std::islower' conflicts with a previous declaration
 2639 |     islower(_CharT __c, const locale& __loc)
Moreover, guarding headers behind #ifdef INCLUDE_STRING inside system.h is insufficient because several C front-end source files pull <locale> transitively without explicitly defining INCLUDE_STRING.

It seems fixed for macOS builds but the same patch not applied for Linux hosts. Before than merge it has to be fixed.

Now, we are applying for linux as well. Please, try again
humbertodias@64e1341

humbertodias · 2026-09-02T14:39:59Z

Thanks for your work @humbertodias . Several of us maintainers greatly appreciate the work you've done for upgrading gcc on PS3 and for coming back to update us. I hope void's reply did not dissuade you.

Do you intend to do any further work on this branch? Or should we consider this ready for review?

I think so. At least, it compiles on both Linux and macOS, and I was able to build a Hello World with ps3dev and run it successfully on RPCS3

humbertodias · 2026-09-02T15:59:49Z

A question: Which commits are the patches generated from? I don't see corresponding pull requests to https://github.com/ps3dev/gcc-PS3 et al. If new base repos are used (which might make sense given a large jump in version), it would be good if those could be moved to the ps3dev org for maintainability.

Got it. Here ps3dev/gcc-PS3#1

Cruslan · 2026-09-02T16:53:52Z

As i tested again on modern host Linux environment (tested on GCC 162.0) uncovered a few compilation failures that prevent the toolchain from building.

1. `libcody/client.cc`: Ambiguous Overload Error on Modern C++ Compilers

When compiling GCC 13.2.0 (scripts/002-gcc-newlib-PPU.sh), compilation of libcody fails:

../../libcody/client.cc: In function 'Cody::Packet Cody::IncludeTranslateResponse(std::vector<std::__cxx11::basic_string<char> >&)':
../../libcody/client.cc:325:42: error: call of overloaded 'Packet(Cody::Client::PacketCode, int)' is ambiguous
  325 |         return Packet (Client::PC_BOOL, 0);
      |                                          ^
note: candidate 1: 'Cody::Packet::Packet(unsigned int, size_t)'
note: candidate 2: 'Cody::Packet::Packet(unsigned int, const char8_t*)'

PR #121 introduced the constructor Packet (unsigned c, const char8_t *s) to libcody/cody.hh under #if __cpp_char8_t >= 201811. However, cody.hh already defines Packet (unsigned c, size_t i = 0). In C++, the integer literal 0 is both a numeric integer and a valid null pointer constant. Under C++20 / C++23 mode on newer GCC host compilers, calling Packet(Client::PC_BOOL, 0) produces an ambiguous overload between size_t and const char8_t*. Integer literals in client.cc has to be cast explicitly to size_t. Update patches/gcc-13.2.0-PS3-host.patch as follows:

--- a/libcody/client.cc
+++ b/libcody/client.cc
@@ -322,9 +322,9 @@ Packet IncludeTranslateResponse (std::vector<std::string> &words)
   if (words[0] == (const char *) u8"BOOL" && words.size () == 2)
     {
       if (words[1] == (const char *) u8"FALSE")
-	return Packet (Client::PC_BOOL, 0);
+	return Packet (Client::PC_BOOL, size_t (0));
       else if (words[1] == (const char *) u8"TRUE")
-	return Packet (Client::PC_BOOL, 1);
+	return Packet (Client::PC_BOOL, size_t (1));
       else
 	return Packet (Client::PC_ERROR, u8"");
     }

2. GDB 8.3.1 Bundled Readline Build Failure Under Host GCC 15/16 (C23 Default)

When building GDB 8.3.1 (scripts/003-gdb-PPU.sh and scripts/007-gdb-SPU.sh), the bundled readline fails during compilation of signals.c:

../../readline/signals.c:385:40: error: passing argument 2 of 'rl_set_sighandler' from incompatible pointer type [-Wincompatible-pointer-types]
  385 |       oh = rl_set_sighandler (SIGALRM, rl_signal_handler, &old_alrm);
../../readline/signals.c:62:36: error: 'return' with a value, in function returning void [-Wreturn-mismatch]
   62 | #  define SIGHANDLER_RETURN return (0)

Recent host GCC compilers default to C23 (-std=gnu23). In C23, empty function declarations like void (*signal ()) (); now mean void (*signal (void)) (void) instead of unprototyped functions with unspecified arguments. Readline's configure script uses void (*signal ()) (); in its compile check for VOID_SIGHANDLER. This fails against glibc's <signal.h> prototype, causing VOID_SIGHANDLER to remain undefined. Consequently, readline expects signal handlers to return int, which triggers -Wreturn-mismatch and -Wincompatible-pointer-types (both errors by default in GCC 14+). Explicitly pass -std=gnu17 and compatibility flags to CFLAGS and CXXFLAGS in both GDB build scripts (scripts/003-gdb-PPU.sh and scripts/007-gdb-SPU.sh):

--- a/scripts/003-gdb-PPU.sh
+++ b/scripts/003-gdb-PPU.sh
@@ -37,6 +37,8 @@ cd ${GDB}/build-ppu
 ## Configure the build.
 ## pyenv ships Python 3.10; gdb 8.3.1 cannot build against that ABI.
+CFLAGS="${CFLAGS:-} -std=gnu17 -Wno-incompatible-pointer-types -Wno-int-conversion" \
+CXXFLAGS="${CXXFLAGS:-} -std=gnu++17 -Wno-narrowing" \
 ../configure --prefix="$PS3DEV/ppu" --target="powerpc64-ps3-elf" \
     --disable-multilib \
     --disable-nls \

--- a/scripts/007-gdb-SPU.sh
+++ b/scripts/007-gdb-SPU.sh
@@ -37,6 +37,8 @@ cd ${GDB}/build-spu
 ## Configure the build.
 ## pyenv ships Python 3.10; gdb 8.3.1 cannot build against that ABI.
+CFLAGS="${CFLAGS:-} -std=gnu17 -Wno-incompatible-pointer-types -Wno-int-conversion" \
+CXXFLAGS="${CXXFLAGS:-} -std=gnu++17 -Wno-narrowing" \
 ../configure --prefix="$PS3DEV/spu" --target="spu" \
     --disable-nls \
     --disable-sim \

3. SPU Binutils 2.22 Compatibility with C23 Host Compilers

Compiling Binutils 2.22 for SPU (scripts/005-binutils-SPU.sh) encounters failures on newer host toolchains due to obsolete K&R style definitions, implicit conversions, and strict C23 compiler checks. Enforce C17 semantics by passing -std=gnu17 in scripts/005-binutils-SPU.sh:

--- a/scripts/005-binutils-SPU.sh
+++ b/scripts/005-binutils-SPU.sh
@@ -37,6 +37,7 @@ cd ${BINUTILS}/build-spu
 ## Configure the build.
+CFLAGS="${CFLAGS:-} -std=gnu17 -Wno-incompatible-pointer-types -Wno-int-conversion -Wno-implicit-function-declaration" \
 ../configure --prefix="$PS3DEV/spu" --target="spu" \
     --disable-nls \
     --disable-shared \

…rn compilers.

humbertodias · 2026-09-02T18:05:58Z

As i tested again on modern host Linux environment (tested on GCC 162.0) uncovered a few compilation failures that prevent the toolchain from building.

1. libcody/client.cc: Ambiguous Overload Error on Modern C++ Compilers

When compiling GCC 13.2.0 (scripts/002-gcc-newlib-PPU.sh), compilation of libcody fails:
../../libcody/client.cc: In function 'Cody::Packet Cody::IncludeTranslateResponse(std::vector<std::__cxx11::basic_string<char> >&)':
../../libcody/client.cc:325:42: error: call of overloaded 'Packet(Cody::Client::PacketCode, int)' is ambiguous
  325 |         return Packet (Client::PC_BOOL, 0);
      |                                          ^
note: candidate 1: 'Cody::Packet::Packet(unsigned int, size_t)'
note: candidate 2: 'Cody::Packet::Packet(unsigned int, const char8_t*)'
PR #121 introduced the constructor Packet (unsigned c, const char8_t *s) to libcody/cody.hh under #if __cpp_char8_t >= 201811. However, cody.hh already defines Packet (unsigned c, size_t i = 0). In C++, the integer literal 0 is both a numeric integer and a valid null pointer constant. Under C++20 / C++23 mode on newer GCC host compilers, calling Packet(Client::PC_BOOL, 0) produces an ambiguous overload between size_t and const char8_t*. Integer literals in client.cc has to be cast explicitly to size_t. Update patches/gcc-13.2.0-PS3-host.patch as follows:
--- a/libcody/client.cc
+++ b/libcody/client.cc
@@ -322,9 +322,9 @@ Packet IncludeTranslateResponse (std::vector<std::string> &words)
   if (words[0] == (const char *) u8"BOOL" && words.size () == 2)
     {
       if (words[1] == (const char *) u8"FALSE")
-	return Packet (Client::PC_BOOL, 0);
+	return Packet (Client::PC_BOOL, size_t (0));
       else if (words[1] == (const char *) u8"TRUE")
-	return Packet (Client::PC_BOOL, 1);
+	return Packet (Client::PC_BOOL, size_t (1));
       else
 	return Packet (Client::PC_ERROR, u8"");
     }
2. GDB 8.3.1 Bundled Readline Build Failure Under Host GCC 15/16 (C23 Default)

When building GDB 8.3.1 (scripts/003-gdb-PPU.sh and scripts/007-gdb-SPU.sh), the bundled readline fails during compilation of signals.c:
../../readline/signals.c:385:40: error: passing argument 2 of 'rl_set_sighandler' from incompatible pointer type [-Wincompatible-pointer-types]
  385 |       oh = rl_set_sighandler (SIGALRM, rl_signal_handler, &old_alrm);
../../readline/signals.c:62:36: error: 'return' with a value, in function returning void [-Wreturn-mismatch]
   62 | #  define SIGHANDLER_RETURN return (0)
Recent host GCC compilers default to C23 (-std=gnu23). In C23, empty function declarations like void (*signal ()) (); now mean void (*signal (void)) (void) instead of unprototyped functions with unspecified arguments. Readline's configure script uses void (*signal ()) (); in its compile check for VOID_SIGHANDLER. This fails against glibc's <signal.h> prototype, causing VOID_SIGHANDLER to remain undefined. Consequently, readline expects signal handlers to return int, which triggers -Wreturn-mismatch and -Wincompatible-pointer-types (both errors by default in GCC 14+). Explicitly pass -std=gnu17 and compatibility flags to CFLAGS and CXXFLAGS in both GDB build scripts (scripts/003-gdb-PPU.sh and scripts/007-gdb-SPU.sh):
--- a/scripts/003-gdb-PPU.sh
+++ b/scripts/003-gdb-PPU.sh
@@ -37,6 +37,8 @@ cd ${GDB}/build-ppu
 ## Configure the build.
 ## pyenv ships Python 3.10; gdb 8.3.1 cannot build against that ABI.
+CFLAGS="${CFLAGS:-} -std=gnu17 -Wno-incompatible-pointer-types -Wno-int-conversion" \
+CXXFLAGS="${CXXFLAGS:-} -std=gnu++17 -Wno-narrowing" \
 ../configure --prefix="$PS3DEV/ppu" --target="powerpc64-ps3-elf" \
     --disable-multilib \
     --disable-nls \
--- a/scripts/007-gdb-SPU.sh
+++ b/scripts/007-gdb-SPU.sh
@@ -37,6 +37,8 @@ cd ${GDB}/build-spu
 ## Configure the build.
 ## pyenv ships Python 3.10; gdb 8.3.1 cannot build against that ABI.
+CFLAGS="${CFLAGS:-} -std=gnu17 -Wno-incompatible-pointer-types -Wno-int-conversion" \
+CXXFLAGS="${CXXFLAGS:-} -std=gnu++17 -Wno-narrowing" \
 ../configure --prefix="$PS3DEV/spu" --target="spu" \
     --disable-nls \
     --disable-sim \
3. SPU Binutils 2.22 Compatibility with C23 Host Compilers

Compiling Binutils 2.22 for SPU (scripts/005-binutils-SPU.sh) encounters failures on newer host toolchains due to obsolete K&R style definitions, implicit conversions, and strict C23 compiler checks. Enforce C17 semantics by passing -std=gnu17 in scripts/005-binutils-SPU.sh:
--- a/scripts/005-binutils-SPU.sh
+++ b/scripts/005-binutils-SPU.sh
@@ -37,6 +37,7 @@ cd ${BINUTILS}/build-spu
 ## Configure the build.
+CFLAGS="${CFLAGS:-} -std=gnu17 -Wno-incompatible-pointer-types -Wno-int-conversion -Wno-implicit-function-declaration" \
 ../configure --prefix="$PS3DEV/spu" --target="spu" \
     --disable-nls \
     --disable-shared \

No problem. Now we are using gcc 16.2 to compile.
humbertodias@a27ba07
Thanks

Cruslan · 2026-09-02T21:03:22Z

As i tested it can compile my PS3 Moonlight project without change single of code. It works like charm. It improved altivec SIMD register usage by huge margin, it will increase performance of homebrews by huge margin. Good work.

Architectural Metric (-O2 Baseline)	GCC 7.2.0 (Legacy Toolchain)	GCC 13.2.0 (Modern Toolchain)	Delta / Variance	Microarchitectural Analysis & Hardware Impact
Total Instruction Count	389,209 insns	396,141 insns	+6,932 (+1.78%)	Controlled loop unrolling and SIMD loop prologue/epilogue peeling.
AltiVec SIMD Vector Ops	57 insns	129 insns	+72 (+126.3%)	Opus audio DSP kernels auto-vectorized into 128-bit VMX `vmaddfp` operations.
Hardware Counter Loops (`bdnz`)	2,070 loops	2,337 loops	+267 (+12.9%)	Counted loops routed to Count Register (CTR) with zero branch penalty.
Redundant Sign-Extends (`extsw`)	9,048 insns	7,435 insns	-1,613 (-17.8%)	Advanced VRP eliminates 1,613 dead cycles on the in-order PPE pipeline.
Condition Compares (`cmpdi`)	9,055 insns	7,806 insns	-1,249 (-13.8%)	Combined comparisons with single-cycle record-form instructions (`mr.`).
Video Receiver Memory Traffic	1,122 ops	1,006 ops	-116 ops (-10.3%)	67 fewer loads and 49 fewer stores in `VideoReceiveThreadProc` stream path.
Total Function Boundaries	7,472 funcs	7,352 funcs	-120 funcs	Inter-procedural whole-program inlining eliminating function call overhead.
Procedure Descriptors (`.opd`)	89,784 B (3,741 entries)	88,344 B (3,681 entries)	-1,440 B (-60 entries)	60 fewer external function descriptors saving 24 bytes per descriptor table.
Stripped Binary Footprint	2,398,848 B (2.39 MB)	2,463,752 B (2.46 MB)	+64.9 KB (+2.71%)	70.3% of size delta is DWARF 5 Call Frame Information (`.eh_frame`) unwinding.
Deployable Package (`.pkg`)	843,264 B (824 KB)	886,208 B (866 KB)	+42.9 KB (+5.09%)	Hardware-ready retail signed NPDRM installation package footprint.

Cruslan · 2026-09-02T21:08:14Z

Works nice.

zeldin · 2026-09-02T22:35:27Z

Got it. Here ps3dev/gcc-PS3#1

Thanks. However, please don't slap a new upstreams version on the master branch; we have an upstream branch where the gcc sources are imported without any PS3 modifications (so exactly what comes in the upstream tarball), and then we merge that into master to keep changes with traceability. The patch for ps3toolchain is then simply generated by git diff upstream master.

You don't need to make a separate pull request for the upstream branch; just keep the same structure in your own repo and submit the master branch commit with the merge intact and we can work with that.

If the new gcc version has diverged too much for a merge to be practical, it's better to create a new repo (still with an upstream branch in the same way) and cherry-pick the commits from the old repo that are still relevant.

This may all seem like extra work, but it does ensure maintainability going forward.

humbertodias · 2026-09-03T00:15:33Z

Got it. Here ps3dev/gcc-PS3#1

Thanks. However, please don't slap a new upstreams version on the master branch; we have an upstream branch where the gcc sources are imported without any PS3 modifications (so exactly what comes in the upstream tarball), and then we merge that into master to keep changes with traceability. The patch for ps3toolchain is then simply generated by git diff upstream master.

You don't need to make a separate pull request for the upstream branch; just keep the same structure in your own repo and submit the master branch commit with the merge intact and we can work with that.

If the new gcc version has diverged too much for a merge to be practical, it's better to create a new repo (still with an upstream branch in the same way) and cherry-pick the commits from the old repo that are still relevant.

This may all seem like extra work, but it does ensure maintainability going forward.

Thanks for the guidance - I've restructured this PR to match that workflow.

upstream on my fork is a clean import of gcc-13.2.0.tar.xz (tag gcc-13.2.0-upstream), with no PS3 changes.
This branch merges that tag into master (Merge tag 'gcc-13.2.0-upstream') and keeps the PS3-only commits on top. There is no separate PR for upstream.

The ps3toolchain patch can be generated with:

git diff upstream gcc-ppu-spu-13.2.0

Please, let me know if i missed anything.

zeldin · 2026-09-04T16:24:51Z

Perfect! Thank you. Now it's also easier to review.

Cruslan · 2026-09-06T12:31:59Z

I wanna see this update, GCC's improved auto-vectorizer and SNS modules significantly improves performance.

Fewnity · 2026-09-06T12:47:05Z

GCM samples from PSL1GHT should be compiled with the update because if I remember correctly, GCM stuff wasn't working well when I tested when the PR was created.

Cruslan · 2026-09-06T12:50:27Z

On my PS3 Moonlight project i compiled pr for myself. I didn't see any problems with that, my project utilized Tiny3D for display.

humbertodias · 2026-09-06T20:57:13Z

Tested with SoRR

bennug64-port-ps3.mp4

humbertodias added 6 commits March 4, 2024 22:16

add gcc-ppu-spu v13.2.0

1a930c0

ignore mac .DS_Store

868e7f8

[fix] github actions

d044950

[gha] python dependency name fixed for ubuntu-latest

b237a9a

[gha] install python2 manually

57d3d7a

[gha] sudo required

de37a37

humbertodias changed the title ~~add gcc-ppu-spu v13.2.0~~ Update GCC to 13.2.0 (PUU-SPU) Mar 5, 2024

humbertodias added 5 commits March 4, 2024 23:18

[gha] add g++ with c11 support

eac7e4f

specific patch file for gcc ppu

2f2633f

[gha] unnecessary -get

afb604f

gcc 9.5.0 for spu

66979e9

ppu only for c lang

3445df8

humbertodias changed the title ~~Update GCC to 13.2.0 (PUU-SPU)~~ Update GCC version (PUU-SPU) Mar 5, 2024

humbertodias changed the title ~~Update GCC version (PUU-SPU)~~ Update GCC version (PPU-SPU) Mar 5, 2024

humbertodias added 4 commits March 5, 2024 02:24

[ppu] add c++

319554b

[ppu] fix for _GLIBCXX_HAVE_STRUCT_DIRENT_D_TYPE=0

c48e4e1

[ppu] HAVE_STRUCT_DIRENT_D_TYPE=0

6c44881

[ppu] HAVE_STRUCT_DIRENT_D_TYPE=0

5af633b

spu-gcc 7.20

5e818ab

humbertodias added 9 commits March 5, 2024 16:52

add c++ lang for spu-gcc

4a6b6f7

alias sed='gsed' for mac

3be7a7d

gsed alias for macOS

e489a56

add python2-dev and libssl-dev for psl1ght build

4fc062a

[gha] only ubuntu for while

9cb47c7

[gha] set default shell to bash

b26dbb0

[gha] use ~/.bashrc

10609f0

[gha] mac already have python installed

2a64760

[gha] get current arch

1e96642

humbertodias added 6 commits August 31, 2026 12:22

Drop zlib fdopen hunk from binutils 2.22 SPU patch

bf78b45

Xcode 16.4 on Intel hits the same isupper/islower

5fe807c

Fix malformed Darwin GCC system.h patches so GNU patch can apply them.

4a556cf

Refresh newlib config.sub so Apple Silicon Darwin is a valid host.

53900d0

Fix malformed GCC 9.5.0 Apple Silicon host-darwin.c hunk

97d977d

Move the Intel macOS CI runner from 15 to 26.

9a5a051

humbertodias changed the title ~~Upgrade GCC version (PPU-SPU)~~ Upgrade GCC 13.2.0 version (PPU-SPU) Sep 1, 2026

humbertodias had a problem deploying to DockerHub September 1, 2026 15:45 — with GitHub Actions Failure

Apply the safe-ctype and libcody host fixes on every OS, not only Darwin

64e1341

Build with host GCC 16.2 in Docker and fix C23/C++20 failures on mode…

a27ba07

…rn compilers.

Stay on Debian 12 so the host-linked cross gcc needs only glibc 2.36

2390d77

humbertodias had a problem deploying to DockerHub September 6, 2026 12:59 — with GitHub Actions Failure

Uh oh!

Conversation

humbertodias commented Mar 5, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Features

Result

Uh oh!

Fewnity commented Mar 5, 2024

Uh oh!

humbertodias commented Mar 5, 2024

Uh oh!

void2012 commented Sep 1, 2026

Uh oh!

TheMrIron2 commented Sep 1, 2026

Uh oh!

zeldin commented Sep 1, 2026

Uh oh!

Cruslan commented Sep 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

1. libcody UTF-8 / char8_t type mismatch

2. safe-ctype.h macro poisoning in gcc/system.h

Uh oh!

humbertodias commented Sep 2, 2026

1. libcody UTF-8 / char8_t type mismatch

2. safe-ctype.h macro poisoning in gcc/system.h

Uh oh!

humbertodias commented Sep 2, 2026

Uh oh!

humbertodias commented Sep 2, 2026

Uh oh!

Cruslan commented Sep 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

1. libcody/client.cc: Ambiguous Overload Error on Modern C++ Compilers

2. GDB 8.3.1 Bundled Readline Build Failure Under Host GCC 15/16 (C23 Default)

3. SPU Binutils 2.22 Compatibility with C23 Host Compilers

Uh oh!

humbertodias commented Sep 2, 2026

1. libcody/client.cc: Ambiguous Overload Error on Modern C++ Compilers

2. GDB 8.3.1 Bundled Readline Build Failure Under Host GCC 15/16 (C23 Default)

3. SPU Binutils 2.22 Compatibility with C23 Host Compilers

Uh oh!

Cruslan commented Sep 2, 2026

Uh oh!

Cruslan commented Sep 2, 2026

Uh oh!

zeldin commented Sep 2, 2026

Uh oh!

humbertodias commented Sep 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

zeldin commented Sep 4, 2026

Uh oh!

Cruslan commented Sep 6, 2026

Uh oh!

Fewnity commented Sep 6, 2026

Uh oh!

Cruslan commented Sep 6, 2026

Uh oh!

humbertodias commented Sep 6, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

humbertodias commented Mar 5, 2024 •

edited

Loading

Cruslan commented Sep 2, 2026 •

edited

Loading

1. `libcody` UTF-8 / char8_t type mismatch

2. `safe-ctype.h` macro poisoning in `gcc/system.h`

1. `libcody` UTF-8 / char8_t type mismatch

2. `safe-ctype.h` macro poisoning in `gcc/system.h`

Cruslan commented Sep 2, 2026 •

edited

Loading

1. `libcody/client.cc`: Ambiguous Overload Error on Modern C++ Compilers

1. `libcody/client.cc`: Ambiguous Overload Error on Modern C++ Compilers

humbertodias commented Sep 3, 2026 •

edited

Loading