Monday
I would like to report a recurring, reproducible crash in Archicad
Every crash terminates with signal SIGBUS (0x0A), code BUS_ADRALN (0x01) - a misaligned memory access - at an address inside dyld itself. We disassembled the raw instruction bytes captured in Archicad's own local GSReport crash cache and confirmed the fault occurs inside dyld's Mach-O header / magic-number validation code, where it reads a 4-byte value from a pointer held in register x0 and compares it against the Mach-O magic numbers 0xfeedfacf / 0xfeedface.
I see two consistent variants of this crash, both present across both machines:
Variant A (most common): the faulting instruction is "ldr w8, [x0]", immediately followed by the magic-number comparison.
Variant B (occasional): the faulting instruction is "ldr w8, [x0, #0x14]", which then loads and checks the magic number at offset 0.
In every occurrence of a given variant, the surrounding instruction bytes are byte-for-byte identical across machines and across different application sessions; only the absolute load address differs, which is consistent with normal dyld shared-cache ASLR slide between boot sessions rather than a different underlying fault. This strongly suggests a single, deterministic bug rather than random memory corruption.
I'm attaching the paired .rpt / .rpta FATAL reports (from GSReport's local cache at ~/Library/Application Support/GRAPHISOFT/BugReporting-29/) for the clearest examples of both variants, spanning both affected machines. We have additional reports available on request - well over a dozen in total.
Monday - last edited Monday
Is this just to alert the rest of us to this potential issue, or have you submitted this to technical support? This Community is not a tech support channel as you know.
Do any particular actions result in this crash? What version of macOS? (Your signature lists an out of date version and this sounds like a macOS issue.)
yesterday
Hi Karl,
I haven't submitted to techical support, as I didn't find any suitable category...
I was hoping someone from GS would see this and open a ticket
yesterday
If you don't see 'Product Support' on that page, then use the 'Local Support' link at the bottom.
Here in Australia/New Zealand, we only just got 'Product Support' last week.
I will ask at the next moderators meeting in a couple of days, if everyone is supposed to see 'Product Support'.
Barry.
yesterday
I have forwarded this bug to GS Community Tech Support.