Fix illumox-x64 build with gcc and clang by am11 · Pull Request #129322 · dotnet/runtime

am11 · 2026-06-12T12:47:04Z

First, our default build is failing with solaris gcc 13, so couple of fixes for those are included (jit, gcenv etc.). Then it includes clang fixes.

For the first time, cross building runtime repo for illumos-x64 with llvm-toolchain has succeeded. It has two issues which I've reported upstream with minimal repro and workarounds:

llvm assembler (llvm-mc): Crash with noexecstack on --target=x86_64-pc-solaris2.11 llvm/llvm-project#202953
LLVM linker (lld): Making fuse-ld=lld work for --target=x86_64-pc-solaris2.11 llvm/llvm-project#202963

For the first one I am adding a workaround in configurecompiler.cmake (if (CMAKE_C_COMPILER_ID MATCHES "GNU") condition). For the second one the hack goes in toolchain.cmake which I am skipping from this PR as I'm running trials to slim it down a bit (current state: https://github.com/am11/runtime/blob/67046c0cf095c5d7888c797226fe736156b5e04c/eng/common/cross/toolchain.cmake#L227-L307).

Together the clang build succeeds on Azure Linux cross environment same as rest of the platforms (freebsd, openbsd, haiku).

Note that the gcc build requires another fix upstream dotnet/dotnet#7206, which I recently broke while fixing sccsid which got broken after we started to use --gc-sections linker option liberally.

dotnet-policy-service · 2026-06-12T12:48:30Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

am11 · 2026-06-12T12:51:09Z

cc @gwr, @AustinWise, I know you guys were looking for clang-based crossbuild at some point. While working on OpenBSD bringup, I figured some internal things out for illumos.

If someone could take care of llvm side of patching, that would help. 🙂

am11 · 2026-06-12T13:00:08Z

        memset(&sa_default, 0, sizeof(sa_default)); // On some architectures, sa_mask is a struct so assigning zero to it doesn't compile
+    #pragma clang diagnostic push
+    #pragma clang diagnostic ignored "-Wstrict-prototypes"
        sa_default.sa_handler = SIG_DFL;


These are raised from system headers; I couldn't find a way to make clang happy, so I went with the suppression (and gcc surprisingly ignores it so we don't need additional #ifdef __clang type of condition).

akoeplinger · 2026-06-12T14:15:25Z

build failure looks related

janvorli · 2026-06-12T16:23:54Z

 #endif

-#ifdef MADV_FREE
+#if defined(MADV_FREE) && !defined(TARGET_SUNOS)


Why is this needed? It is strange that it would have MADV_FREE defined and it would fail in the code below.

Mono has HAVE_MADVISE, coreclr doesn't. The error was about madvise not being defined:

[117/162] Building CXX object nativeaot/Runtime/Full/CMakeFiles/Runtime.ServerGC.dir/__/__/__/gc/unix/gcenv.unix.cpp.o FAILED: nativeaot/Runtime/Full/CMakeFiles/Runtime.ServerGC.dir/__/__/__/gc/unix/gcenv.unix.cpp.o /crossrootfs/x64/bin/x86_64-illumos-g++ --sysroot=/crossrootfs/x64 -DDISABLE_CONTRACTS -DFEATURE_BASICFREEZE -DFEATURE_CONSERVATIVE_GC -DFEATURE_EVENT_TRACE -DFEATURE_HIJACK -DFEATURE_MANUALLY_MANAGED_CARD_BUNDLES -DFEATURE_NATIVEAOT -DFEATURE_PERFTRACING -DFEATURE_READONLY_GS_COOKIE -DFEATURE_RX_THUNKS -DFEATURE_SVR_GC -DFEATURE_USE_SOFTWARE_WRITE_WATCH_FOR_GC_HEAP -DGC_DESCRIPTOR -DHOST_64BIT -DHOST_AMD64 -DHOST_UNIX -DNATIVEAOT -DNDEBUG -DTARGET_64BIT -DTARGET_AMD64 -DTARGET_ILLUMOS -DTARGET_SUNOS -DTARGET_UNIX -DUNIX_AMD64_ABI -DURTBLDENV_FRIENDLY=Retail -DVERIFY_HEAP -D_FILE_OFFSET_BITS=64 -D_LIB -D_LIBUNWIND_DISABLE_ZERO_COST_APIS=1 -D_LIBUNWIND_IS_NATIVE_ONLY -D_POSIX_PTHREAD_SEMANTICS -D_REENTRANT -D_TIME_BITS=64 -D_XPG4_2 -D__EXTENSIONS__ -I/runtime/artifacts/obj/coreclr/illumos.x64.Release/nativeaot/Runtime/Full -I/runtime/src/coreclr/nativeaot/Runtime/Full -I/runtime/src/native/external/llvm-libunwind/include -I/runtime/src/native -I/runtime/src/native/inc -I/runtime/src/coreclr/pal/prebuilt/inc -I/runtime/artifacts/obj -I/runtime/src/coreclr/nativeaot/Runtime/inc -I/runtime/src/coreclr/nativeaot/Runtime/. -I/runtime/src/coreclr/nativeaot/Runtime/../../gc -I/runtime/src/coreclr/nativeaot/Runtime/../../gc/env -I/runtime/artifacts/obj/coreclr/illumos.x64.Release/nativeaot/Runtime/eventpipe/inc -I/runtime/src/coreclr/runtime -I/runtime/src/coreclr/nativeaot/Runtime/unix -I/runtime/src/coreclr/nativeaot/Runtime/../../pal/inc/rt -I/runtime/src/coreclr/nativeaot/Runtime/amd64 -I/runtime/artifacts/obj/coreclr/illumos.x64.Release/nativeaot/Runtime -I/runtime/src/coreclr/debug/datadescriptor-shared/inc -isystem /crossrootfs/x64/include -fstack-protector -O3 -DNDEBUG -std=gnu++17 -fPIC -O3 -Wall -g -fno-omit-frame-pointer -fno-strict-overflow -fno-strict-aliasing -fstack-protector-strong -Werror -Wno-unused-variable -Wno-unused-value -Wno-unused-function -Wno-tautological-compare -Wno-unknown-pragmas -Wimplicit-fallthrough -Wvla -Wno-invalid-offsetof -Wno-unused-but-set-variable -ffp-contract=off -fno-rtti -Wno-uninitialized -Wno-strict-aliasing -Wno-array-bounds -Wno-misleading-indentation -Wno-stringop-overflow -Wno-restrict -Wno-stringop-truncation -Wno-class-memaccess -faligned-new -fsigned-char -fvisibility=hidden -ffunction-sections -fdata-sections -fno-exceptions -fno-asynchronous-unwind-tables -nostdlib -MD -MT nativeaot/Runtime/Full/CMakeFiles/Runtime.ServerGC.dir/__/__/__/gc/unix/gcenv.unix.cpp.o -MF nativeaot/Runtime/Full/CMakeFiles/Runtime.ServerGC.dir/__/__/__/gc/unix/gcenv.unix.cpp.o.d -o nativeaot/Runtime/Full/CMakeFiles/Runtime.ServerGC.dir/__/__/__/gc/unix/gcenv.unix.cpp.o -c /runtime/src/coreclr/gc/unix/gcenv.unix.cpp /runtime/src/coreclr/gc/unix/gcenv.unix.cpp: In static member function 'static bool GCToOSInterface::VirtualReset(void*, size_t, bool)': /runtime/src/coreclr/gc/unix/gcenv.unix.cpp:574:10: error: 'madvise' was not declared in this scope; did you mean 'raise'? 574 | st = madvise(address, size, MADV_FREE); | ^~~~~~~ | raise [118/162] Building C object /runtime/artifacts/obj/external/libunwind/CMakeFiles/libunwind.dir/runtime/src/native/external/libunwind/src/dwarf/Lparser.c.o [119/162] Building C object /runtime/artifacts/obj/external/libunwind/CMakeFiles/libunwind.dir/runtime/src/native/external/libunwind/src/dwarf/Gexpr.c.o [120/162] Building C object /runtime/artifacts/obj/external/libunwind/CMakeFiles/libunwind.dir/runtime/src/native/external/libunwind/src/dwarf/Gparser.c.o [121/162] Building CXX object Corehost.Static/hostmisc/CMakeFiles/hostmisc.dir/utils.cpp.o [122/162] Building CXX object Corehost.Static/hostmisc/CMakeFiles/hostmisc_public.dir/utils.cpp.o ninja: build stopped: subcommand failed. /runtime/src/coreclr Failed to build "CoreCLR component". /runtime/src/coreclr/runtime.proj(120,5): error MSB3073: The command ""/runtime/src/coreclr/build-runtime.sh" -x64 -release gcc -cross -os illumos -ninja -targetrid illumos-x64 -cmakeargs "-DCLR_DOTNET_RID=illumos-x64" -cmakeargs "-DCLR_DOTNET_HOST_PATH='/runtime/.dotnet/dotnet'" -cmakeargs "-DCDAC_BUILD_TOOL_BINARY_PATH=/runtime/artifacts/bin/coreclr/illumos.x64.Release/cdac-build-tool/cdac-build-tool.dll" -cmakeargs "-DFEATURE_DYNAMIC_CODE_COMPILED=1" " exited with code 1. Build FAILED. /runtime/src/coreclr/runtime.proj(120,5): error MSB3073: The command ""/runtime/src/coreclr/build-runtime.sh" -x64 -release gcc -cross -os illumos -ninja -targetrid illumos-x64 -cmakeargs "-DCLR_DOTNET_RID=illumos-x64" -cmakeargs "-DCLR_DOTNET_HOST_PATH='/runtime/.dotnet/dotnet'" -cmakeargs "-DCDAC_BUILD_TOOL_BINARY_PATH=/runtime/artifacts/bin/coreclr/illumos.x64.Release/cdac-build-tool/cdac-build-tool.dll" -cmakeargs "-DFEATURE_DYNAMIC_CODE_COMPILED=1" " exited with code 1. 0 Warning(s) 1 Error(s)

Interesting, so sunos has MADV_FREE defined and yet no madvise, that's quite weird.

It's in <sys/mman.h>

#if !defined(_STRICT_POSIX) extern int mincore(caddr_t, size_t, char *); extern int memcntl(void *, size_t, int, void *, int, int); extern int madvise(void *, size_t, int); [...]

Does the build supply -D_STRICT_POSIX or something?

It's consistent once you see it's a namespace thing: madvise is a non-standard SVR4/BSD extension, posix_madvise is the standardized one, and we build with -D_XPG4_2 (strict X/Open). From the rootfs's sys/mman.h:

#if (_POSIX_C_SOURCE > 2) || defined(_XPG4_2) extern void *mmap(void *, size_t, int, int, int, off_t); ... #else /* (_POSIX_C_SOURCE > 2) || defined(_XPG4_2) */ ... extern int madvise(caddr_t, size_t, int); ... #endif #if !defined(__XOPEN_OR_POSIX) || defined(_XPG6) || defined(__EXTENSIONS__) extern int posix_madvise(void *, size_t, int); #endif #if (_POSIX_C_SOURCE <= 2) && !defined(_XPG4_2) || defined(__EXTENSIONS__) ... #define MADV_FREE 5 /* contents can be freed */ ... #endif

So on illumos the posix-y path is the preferred one, and posix_madvise(POSIX_MADV_DONTNEED) is what they consider "modern".

jakobbotsch

JIT part LGTM

janvorli · 2026-06-12T20:06:13Z

            {
                assert(origHandler->sa_handler);
-                origHandler->sa_handler(sig);
+                ((void (*)(int))origHandler->sa_handler)(sig);


What is the sa_handler type on illumos?

C and C++ have different types:

struct sigaction { int sa_flags; union { #ifdef __cplusplus void (*_handler)(int); #else void (*_handler)(); #endif #if defined(__EXTENSIONS__) || defined(_KERNEL) || \ (!defined(_STRICT_STDC) && !defined(__XOPEN_OR_POSIX)) || \ (_POSIX_C_SOURCE > 2) || defined(_XPG4_2) void (*_sigaction)(int, siginfo_t *, void *); #endif } _funcptr; sigset_t sa_mask; #ifndef _LP64 int sa_resv[2]; #endif }; #define sa_handler _funcptr._handler #define sa_sigaction _funcptr._sigaction

https://github.com/illumos/illumos-gate/blob/2748bf5666cab238e5b0138f9baea36ee283f990/usr/src/uts/common/sys/signal.h#L88

Our rootfs has an older version of kernel. There is no new update from https://github.com/illumos/sysroot/ in a while.

If you set sa_handler, the right type is (void (*)(int)), and if you
set sa_sigaction, the right type is (void (*)(int, siginfo_t *, void *))
Those are union arms. Here'e the relevant section of sys/signa.h

/* * The signal handler routine can have either one or three arguments. With * K&R C code could use either form so not specifing the arguments neatly * finessed the problem. Modern C and any C++ do not accept this. To them * "(*sa_handler)()" indicates a routine with no arguments (what used to be * "(*sa_handler)(void)"). One or the other form must be used and the only * logical choice is "(*sa_handler)(int)" to allow the SIG_* defines to work. * "(*sa_sigaction)(int, siginfo_t *, void *)" can be used for the three * argument form. */ /* * Note: storage overlap by sa_handler and sa_sigaction */ struct sigaction { int sa_flags; union { void (*_handler)(int); #if defined(__EXTENSIONS__) || defined(_KERNEL) || \ (!defined(_STRICT_STDC) && !defined(__XOPEN_OR_POSIX)) || \ (_POSIX_C_SOURCE > 2) || defined(_XPG4_2) void (*_sigaction)(int, siginfo_t *, void *); #endif } _funcptr; sigset_t sa_mask; #ifndef _LP64 int sa_resv[2]; #endif }; #define sa_handler _funcptr._handler #define sa_sigaction _funcptr._sigaction

Fix illumox-x64 build with gcc and clang

9abbcb4

github-actions Bot added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label Jun 12, 2026

dotnet-policy-service Bot added the community-contribution Indicates that the PR has been added by a community member label Jun 12, 2026

am11 added area-PAL-coreclr only for closed issues os-SunOS SunOS, currently not officially supported and removed area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI labels Jun 12, 2026

am11 requested review from akoeplinger, jakobbotsch and janvorli June 12, 2026 12:51

Apply format.patch

72ba266

am11 commented Jun 12, 2026

View reviewed changes

jakobbotsch reviewed Jun 12, 2026

View reviewed changes

Comment thread src/coreclr/jit/gentree.h Outdated

am11 added 3 commits June 12, 2026 15:27

Address CR feedback

acc3f32

Merge branch 'main' into feature/port/illumos-native-build2

626a07c

Apply format patch

cd79548

janvorli reviewed Jun 12, 2026

View reviewed changes

am11 added 2 commits June 12, 2026 20:18

Fix debug assert annotation

e0346a2

Merge branch 'main' into feature/port/illumos-native-build2

7fe19eb

jakobbotsch approved these changes Jun 12, 2026

View reviewed changes

janvorli reviewed Jun 12, 2026

View reviewed changes

This was referenced Jun 13, 2026

The Operation will be canceled. The next steps may not contain expected logs. dotnet/dnceng#3008

Open

Multiple Helix work items fail on maccatalyst/tvos CoreCLR Release #126460

Open

Arm/Arm64: Segfault in System.Security.Cryptography.Tests #129339

Open

Conversation

am11 commented Jun 12, 2026

Uh oh!

dotnet-policy-service Bot commented Jun 12, 2026

Uh oh!

am11 commented Jun 12, 2026

Uh oh!

am11 Jun 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

akoeplinger commented Jun 12, 2026

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gwr Jun 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jakobbotsch left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

am11 Jun 12, 2026 •

edited

Loading

gwr Jun 12, 2026 •

edited

Loading