Revert "feat(glx-runtime): GL 来自生态而非 /usr/lib"

main is red at f44e896 and the previous run was green, so this is mine.

The failures are not graphics ones -- asio-module and core fail too, all with
the same shape:

    .../xim-x-glibc/2.39/lib64/libc.so.6: version `GLIBC_2.42' not found
                                          (required by <the test binary>)

mesa declares `xim:glibc@>=2.38`, a floor rather than a pin, so pulling
`xim:graphics` in as a dependency installed glibc 2.44 alongside 2.39. mcpp's
payload probe takes the HIGHEST version it finds, so every build in that home
started compiling and linking against 2.44 -- while the gcc payload's baked-in
interpreter and specs still name 2.39. The result is binaries that reference
GLIBC_2.42/2.43 symbols and a runtime that cannot provide them, in packages
that have nothing to do with graphics.

So the dependency did not just add a graphics stack: it silently moved the
glibc every build in the home compiles against, and only the compile half
moved. That is a real defect in how mcpp picks a payload version, not
something to paper over in this recipe -- and it has to be fixed on the mcpp
side before this change can land again.

The GL work itself was verified (payload libEGL, GPU renderer, no C runtime on
the consumer's RUNPATH) and is not in question. What was missing was any check
of what installing the stack does to the REST of a home.

Author: Sunrisepeak

pkgs/c/compat.glfw.lua

@@ -80,7 +80,7 @@ package = {
                 "src/posix_module.c",
             },
             deps = {
-                ["compat.glx-runtime"] = "2026.08.08",
+                ["compat.glx-runtime"] = "2026.06.03",
                 ["compat.x11"]       = "1.8.13",
                 ["compat.xcursor"]   = "1.2.3",
                 ["compat.xext"]      = "1.3.7",

pkgs/c/compat.glx-runtime.lua

@@ -2,59 +2,13 @@ package = {
     spec        = "1",
     namespace   = "compat",
     name        = "glx-runtime",
-    description = "GLVND/GLX/OpenGL runtime for mcpp Linux window applications, from the xlings graphics stack",
+    description = "Host GLVND/GLX/OpenGL runtime adapter for mcpp Linux window applications",
     licenses    = {"MIT"},
     repo        = "https://github.com/KhronosGroup/OpenGL-Registry",
     type        = "package",
 
-    -- WHERE THE GL RUNTIME COMES FROM, AND WHY IT CHANGED
-    --
-    -- Until 2026.08.08 this package symlinked the HOST's libGL/libEGL out of
-    -- /usr/lib*. That is the thing mcpp#352 is: the host's Mesa needs
-    -- GLIBC_2.43 and mcpp's payload glibc is 2.39, so the program linked
-    -- cleanly and exited 255 with no output. It is also the boundary the
-    -- xlings hermetic policy names first -- any .so under /usr/lib* or /lib*.
-    --
-    -- The runtime now comes from `xim:graphics`, the ecosystem's own stack:
-    -- 22 packages plus two sentinels that probe for a host-side userspace half
-    -- they do not own (the proprietary NVIDIA driver, WSL2's D3D12) and
-    -- succeed having linked nothing when it is absent. One dependency, every
-    -- host shape, no conditional in this file.
-    --
-    -- Measured on an NVIDIA host after the change: libEGL resolves to
-    -- xim-x-libglvnd/1.7.0/lib/libEGL.so.1 and GL_RENDERER is the GPU, not
-    -- llvmpipe. Both halves of that matter -- "a window appeared" is a false
-    -- pass, because llvmpipe renders one too.
     xpm = {
         linux = {
-            -- The whole hermetic graphics stack. A RUNTIME dep, not a build
-            -- one: nothing here compiles against it, the produced consumer
-            -- loads it.
-            --
-            -- PLATFORM level, beside the version entries rather than inside
-            -- one. Every other recipe in both indexes places it here, and the
-            -- first attempt at this change put it inside the 2026.08.08 entry:
-            -- the descriptor parsed, the stack was never installed, and the
-            -- install failed on the required-library check -- an error naming
-            -- libGL.so.1 rather than the misplaced key. Whether a per-version
-            -- `deps` is rejected or merely unread was not determined; what is
-            -- established is that it does not take effect.
-            --
-            -- It therefore also applies to the legacy 2026.06.03 entry below,
-            -- which does not use it. That costs a consumer still pinned there
-            -- a download it will not read, and the alternative -- deleting the
-            -- published version -- would break them outright.
-            deps = { runtime = { "xim:graphics" } },
-            ["2026.08.08"] = {
-                url    = {
-                    GLOBAL = "https://raw.githubusercontent.com/KhronosGroup/OpenGL-Registry/a30033d3e812c9bf10094f1010374a6b15e192eb/README.adoc",
-                    CN     = "https://gitcode.com/mcpp-res/glx-runtime/releases/download/2026.08.08/glx-runtime-2026.08.08.adoc",
-                },
-                sha256 = "ea68efce197e68413ebb62c51ab4bccfb2309a2fca776d31b49d972f59f3640e",
-            },
-            -- Kept so already-published consumers pinned to it keep resolving.
-            -- It sources libGL from the HOST and is the configuration behind
-            -- mcpp#352; new consumers must not pin it.
             ["2026.06.03"] = {
                 url    = {
                     GLOBAL = "https://raw.githubusercontent.com/KhronosGroup/OpenGL-Registry/a30033d3e812c9bf10094f1010374a6b15e192eb/README.adoc",
@@ -87,7 +41,6 @@ package = {
 }
 
 import("xim.libxpkg.pkginfo")
-import("xim.libxpkg.system")
 import("xim.libxpkg.log")
 
 local function sh_quote(value)
@@ -107,26 +60,6 @@ local function split_paths(value)
     return out
 end
 
--- Where to take the GL libraries from.
---
--- The SUBOS VIEW (`<subos>/lib`), not a payload directory. A payload path pins
--- a version, so a consumer's recorded RUNPATH would name mesa 25.0.7.1 forever
--- and stop resolving the day it is upgraded; the view is the stable
--- indirection -- the role /run/opengl-driver plays on NixOS. xlings repoints
--- it as the active version changes and this package needs no new release.
---
--- The view also carries libc.so.6, crt1.o and the rest of the C runtime, and
--- those must NEVER reach a consumer's RUNPATH: the consumer runs under mcpp's
--- payload loader, and pairing one loader with another glibc's libc.so.6 faults
--- inside the dynamic linker before main, with empty output. What keeps them
--- out is the pattern list below -- so that list is a safety boundary, not a
--- convenience, and nothing resembling `libc*` may ever be added to it.
---
--- MCPP_HOST_GL_LIBRARY_PATH still works and is now the ONLY door back to the
--- host. Using it leaves the hermetic guarantee: the libraries it names were
--- built against the host's glibc, and loading them under mcpp's payload glibc
--- is exactly the configuration mcpp#352 reports. It exists for a machine whose
--- GPU vendor the ecosystem does not cover yet.
 local function candidate_dirs()
     local out = {}
     local seen = {}
@@ -138,12 +71,13 @@ local function candidate_dirs()
     end
 
     for _, dir in ipairs(split_paths(os.getenv("MCPP_HOST_GL_LIBRARY_PATH"))) do
-        log.warn("MCPP_HOST_GL_LIBRARY_PATH names %s: GL will come from the "
-                 .. "host, which is the configuration behind mcpp#352", dir)
         add(dir)
     end
-
-    add(path.join(system.subos_sysrootdir(), "lib"))
+    add("/lib/x86_64-linux-gnu")
+    add("/usr/lib/x86_64-linux-gnu")
+    add("/lib64")
+    add("/usr/lib64")
+    add("/usr/lib")
     return out
 end
 
@@ -156,11 +90,7 @@ local host_gl_patterns = {
     "libEGL.so*",
     "libEGL_*.so*",
     "libGLES*.so*",
-    -- No libnvidia* here. The proprietary driver reaches the subos through
-    -- xim:nvidia-gl-host-link, which links it under the glvnd vendor names
-    -- already matched above; taking it by its own name would be a second
-    -- route to the same libraries, and the two would disagree the day the
-    -- driver is upgraded under us.
+    "libnvidia*.so*",
     "libglapi.so*",
     "libdrm*.so*",
     "libexpat.so*",
@@ -174,43 +104,90 @@ local required = {
     ["libGL.so.1"] = false,
 }
 
+-- Is FILE a 64-bit ELF? e_ident[EI_CLASS] == ELFCLASS64.
+--
+-- Five bytes read directly. `file`/`readelf`/`patchelf` would each answer this
+-- and each may be absent when a hook runs, and a probe that answers "cannot
+-- tell" by assuming "fine" is the bug below.
+local function is_elf64(file)
+    local f = io.open(file, "rb")
+    if not f then return false end
+    local head = f:read(5)
+    f:close()
+    return head ~= nil and #head == 5
+       and head:sub(1, 4) == "\127ELF" and head:byte(5) == 2
+end
+
+-- Link the host's GL runtime into one directory, FIRST HIT WINS, 64-bit only.
+--
+-- openxlings/xlings' mcpp#352: on Fedora 44 this produced
+--     libGLX.so.0 -> /usr/lib/libGLX.so.0
+-- a 32-bit library, and the application died with
+--     libGLX.so.0: wrong ELF class: ELFCLASS32
+--
+-- TWO BUGS, and the obvious diagnosis ("the candidate order assumes Debian") is
+-- not either of them -- `/usr/lib64` is already ahead of `/usr/lib` in the list:
+--
+--   1. `ln -sf` OVERWRITES. The loop reached /usr/lib64 first and linked the
+--      correct file, then reached /usr/lib and replaced it. Last-wins, not
+--      first-wins. `libOpenGL.so.0` survived as 64-bit purely because that host's
+--      32-bit glvnd does not ship it -- which is why exactly one link in the bug
+--      report was right.
+--   2. NO ABI CHECK ANYWHERE, including in `required` below, which asserted that
+--      libGLX.so.0 and libGL.so.1 EXIST. Both existed. Both were 32-bit.
+--
+-- There is no directory layout to assume: the FHS biarch clause makes /usr/lib
+-- 32-bit (Fedora/RHEL/SUSE), Debian explicitly declined that clause and uses
+-- /usr/lib/<triplet> so its /usr/lib is 64-bit, and Arch is a third answer
+-- again. So the fix cannot be a better ordering -- it has to be an ABI check,
+-- which makes the order stop mattering.
 local function link_runtime_libs(outdir)
     os.mkdir(outdir)
+    local claimed = {}
     for _, dir in ipairs(candidate_dirs()) do
         for _, pattern in ipairs(host_gl_patterns) do
-            os.exec(
-                "for lib in " .. sh_quote(dir) .. "/" .. pattern ..
-                "; do [ -e \"$lib\" ] || continue; " ..
-                "ln -sf \"$lib\" " .. sh_quote(outdir) .. "/\"$(basename \"$lib\")\"; " ..
-                "done"
-            )
+            -- Enumerate, then decide per file, instead of letting the shell
+            -- link them: the decision needs the ELF class and "have I already
+            -- taken this name", neither of which a `ln -sf` loop can express.
+            local pipe = io.popen("ls -1 " .. sh_quote(dir) .. "/" .. pattern
+                                  .. " 2>/dev/null")
+            if pipe then
+                for line in pipe:lines() do
+                    local lib = line:gsub("[\r\n]+$", "")
+                    local name = lib:match("[^/]+$")
+                    if lib ~= "" and name and not claimed[name]
+                       and is_elf64(lib) then
+                        claimed[name] = lib
+                        os.exec("ln -sf " .. sh_quote(lib) .. " "
+                                .. sh_quote(path.join(outdir, name)))
+                    end
+                end
+                pipe:close()
+            end
         end
     end
 
     for name, _ in pairs(required) do
-        if not os.isfile(path.join(outdir, name)) then
-            log.error("%s is not in this subos. The GL runtime comes from "
-                      .. "`xim:graphics`; if it is declared and this still "
-                      .. "fires, the stack did not finish installing", name)
+        local link = path.join(outdir, name)
+        -- Existence AND ABI. Existence alone passed on the Fedora host with
+        -- both links 32-bit, which is how a broken package reported success and
+        -- the failure surfaced as a silent exit code 255 from the application.
+        if not os.isfile(link) then
+            log.error("required host GL runtime library not found: %s", name)
+            log.error("  searched: %s", table.concat(candidate_dirs(), " "))
+            log.error("  install your distro's GL runtime (mesa / libglvnd)")
             return false
         end
-    end
-
-    -- Nothing resembling a C runtime may have come along. Asserted rather
-    -- than trusted: the pattern list is what keeps it out, and a pattern is
-    -- one careless edit away from matching more than it meant to. The failure
-    -- it prevents has no diagnostic of its own -- the consumer dies inside
-    -- the dynamic linker before main, printing nothing.
-    for _, bad in ipairs({"libc.so.6", "libc.so", "ld-linux-x86-64.so.2",
-                          "libpthread.so.0", "libdl.so.2", "libm.so.6"}) do
-        if os.isfile(path.join(outdir, bad)) then
-            log.error("%s was linked into the GL runtime directory. It would "
-                      .. "land on every consumer's RUNPATH and pair a second "
-                      .. "libc with mcpp's loader, which faults before main "
-                      .. "with no output at all", bad)
+        if not is_elf64(link) then
+            log.error("host %s is not 64-bit (%s)", name, claimed[name] or link)
+            log.error("  a 32-bit library here fails at dlopen with")
+            log.error("  `wrong ELF class: ELFCLASS32` and the application")
+            log.error("  exits without output. Install the 64-bit GL runtime.")
             return false
         end
     end
+    log.info("glx-runtime: linked %d host GL libraries (64-bit)",
+             (function() local n = 0 for _ in pairs(claimed) do n = n + 1 end return n end)())
     return true
 end