GT crashes at startup on macOS 27 (Apple Silicon) — JIT code-zone allocation fails
Summary: On macOS 27.0 (Darwin 27, Apple Silicon), GlamorousToolkit segfaults during VM init while reserving the Cog code zone, before the image loads. Fully reproducible; the image is never reached.
Environment
- macOS 27.0 beta, build
26A5353q (Darwin 27.0.0), Apple M1 Max (arm64)
- GT v1.1.436 — also reproduces on v1.1.417 and v1.1.400
- VM:
5.0-Pharo 10.3.9-46-gea9a2ea, built May 29 2026, Apple LLVM 14.0.0 (clang-1400.0.29.202), Production Spur 64-bit
Crash
EXC_BAD_ACCESS (SIGSEGV), KERN_INVALID_ADDRESS at 0x0, faulting thread is the PharoVM thread, pc = lr = 0 (jump through a null pointer).
- Backtrace (symbolicated):
??? +0
libPharoVMCore.dylib allocateMemoryForImagewithHeader+548
libPharoVMCore.dylib readImageNamed+3840
libPharoVMCore.dylib vm_init+208
GlamorousToolkit vm_bindings::interpreter::PharoInterpreter::run+36
GlamorousToolkit std::sys::backtrace::__rust_begin_short_backtrace+76
GlamorousToolkit core::ops::function::FnOnce::call_once+196
GlamorousToolkit std::sys::thread::unix::Thread::new::thread_start+408
libsystem_pthread.dylib _pthread_start+136
libsystem_pthread.dylib thread_start+8
allocateMemoryForImagewithHeader is allocating the JIT code zone; the disassembly of that routine contains the error string "Could not allocate codeZone in the expected place (%p), got %p", which strongly suggests the code zone can no longer be reserved at the expected virtual address under macOS 27's address-space layout.
Reproduction: Launch GlamorousToolkit on macOS 27 (Apple Silicon). Crashes immediately, every time.
Things ruled out
- Not worker-thread specific —
--worker no reproduces identically.
- Not image corruption — the image-header size is read correctly (matches file size minus header); the crash is at heap/code-zone reservation, not parsing.
- Not quarantine / App Translocation — reproduces from a non-translocated path.
Happy to provide the full .ips privately if useful.
GT crashes at startup on macOS 27 (Apple Silicon) — JIT code-zone allocation fails
Summary: On macOS 27.0 (Darwin 27, Apple Silicon), GlamorousToolkit segfaults during VM init while reserving the Cog code zone, before the image loads. Fully reproducible; the image is never reached.
Environment
26A5353q(Darwin 27.0.0), Apple M1 Max (arm64)5.0-Pharo 10.3.9-46-gea9a2ea, built May 29 2026, Apple LLVM 14.0.0 (clang-1400.0.29.202), Production Spur 64-bitCrash
EXC_BAD_ACCESS (SIGSEGV),KERN_INVALID_ADDRESS at 0x0, faulting thread is thePharoVMthread,pc = lr = 0(jump through a null pointer).allocateMemoryForImagewithHeaderis allocating the JIT code zone; the disassembly of that routine contains the error string "Could not allocate codeZone in the expected place (%p), got %p", which strongly suggests the code zone can no longer be reserved at the expected virtual address under macOS 27's address-space layout.Reproduction: Launch GlamorousToolkit on macOS 27 (Apple Silicon). Crashes immediately, every time.
Things ruled out
--worker noreproduces identically.Happy to provide the full
.ipsprivately if useful.