Arch Linux User Repository

Hi @algebro, I went ahead and made a patch based on PR #1973. You need to remove the tail-end of that patch (the third-party/fmt part). Here's the PKGBUILD diff:

diff --git a/PKGBUILD b/PKGBUILD
index 8d968a6..70a9a53 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -2,7 +2,7 @@

 pkgname=pcsx-redux-git
 _pkgname=pcsx-redux
-pkgver=r6790.718f0912
+pkgver=r6791.47d83a2b
 pkgrel=1
 pkgdesc='Modern fork of the pcsxr PlayStation 1 emulator focused on reverse engineering and homebrew development'
 arch=('x86_64' 'aarch64')
@@ -57,6 +57,7 @@ source=("${_pkgname}::git+https://github.com/grumpycoders/pcsx-redux.git"
         'git+https://github.com/taocpp/PEGTL.git'
         'git+https://github.com/grumpycoders/psxlua.git'
         'pcsx-redux.sh'
+        'fmt12.patch'
         )
 sha256sums=('SKIP'
             'SKIP'
@@ -88,7 +89,8 @@ sha256sums=('SKIP'
             'SKIP'
             'SKIP'
             'SKIP'
-            '21db8ce528f3240388d55d1f309ebc6060bb1145cce50553659e73fb6f89d326')
+            '21db8ce528f3240388d55d1f309ebc6060bb1145cce50553659e73fb6f89d326'
+            '8a3aaf1defa1ac4337cd23ab8917eef24dd0717d5ed565e599b63c69fc1def12')

 pkgver() {
   cd "$_pkgname"
@@ -164,6 +166,7 @@ prepare() {

 build() {
   cd "$_pkgname"
+  cat "$srcdir/fmt12.patch" | patch -p1
   make
 }

Thanks for investigating this @rubin55. For Arch packages it's generally preferred to link against system packages when possible, so using the submodule was a mistake. I've updated the package to use the system fmt and I'll take a stab at fixing the issue upstream since it is technically a bug there and doesn't look too complicated.

I noticed that even though this package pulls an fmt as a git module, it actually links against system libfmt; this must be an error in how the build is done? it is especially problematic now Arch is on fmt 12, which removed fmt::localtime, which pcsx-redux makes use of. I filed https://github.com/grumpycoders/pcsx-redux/issues/1969 upstream.

If this package would actually use the fmt submodule, the fact that arch upgraded to fmt 12 would not matter; is the linking against system fmt intentional?

Thanks for the suggestions @HurricanePootis. How did you get the depends() packages from namcap? I tried running it and I'm not getting it to display anything (other than a missing contributor tag with --info)

Hello, this package could do with the following changes:

1. Add conflicts and provides since this is a -git package
2. Add all depends() listed by namcap
3. Add imagemagick to makedepends() for simplified install step
4. Add missing submodules
5. Simplify the whole git submodule process. I see that you don't want to clone any submodule not required, which can be done by just: 1. configuring the submodules you want; 2. running git submodule update without the --init.
6. pcsx-redux now has an install function in the makefile, so thats what we use
7. Use SDPX compliant license()

Here is the link to the Github gist which contains the patch file.

If you do not want to manually go through this patch file, I do not mind being listed as co-maintainer myself. Thank you for your hardwork!

Any suggestions other than building in a container?

Build in a clean chroot with devtools See https://wiki.archlinux.org/title/DeveloperWiki:Building_in_a_clean_chroot extra-x86_64-build

Seems to use system fmt on my system, resulting in build failure:

/usr/include/fmt/base.h:2248:29: error: passing ‘const fmt::v11::formatter<PCSX::IEC60908b::MSF>’ as ‘this’ argument discards qualifiers [-fpermissive]
 2248 |     ctx.advance_to(cf.format(*static_cast<qualified_type*>(arg), ctx));

Any suggestions other than building in a container?

Very aggressive compilation settings. My session crashes.

@jpyper I went through my notes and figured out that it was the -fstack-protector-strong flag that broke compilation for luajit. Do you know of a good/easy way to negate just that flag but keep all the others? Note that some are set by makepkg.conf and some are set by the Makefile.

@algebro The "AARCH64" in the arch= line should be lowercase "aarch64". :-]

Again, thank you for your work.