[GL]test-exc fails on Ubuntu20.04_ppc

[bregma]

The make check tests Gtest-exc and Ltest-exc fail on ubuntu20.04_ppc (32-bit big-endian) in github actions.

FAIL: Gtest-exc
===============

 >unw_init_local_common: (cursor=0x40ffe6cc)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000d20)
 >_Uppc32_dwarf_search_unwind_table: e->fde_offset = 2ec, segbase = 10010878, debug_frame_base = 0, fde_addr = 10010b64
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000f38)
 >_Uppc32_dwarf_search_unwind_table: e->fde_offset = 334, segbase = 10010878, debug_frame_base = 0, fde_addr = 10010bac
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010001048)
 >_Uppc32_dwarf_search_unwind_table: e->fde_offset = 358, segbase = 10010878, debug_frame_base = 0, fde_addr = 10010bd0
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_dwarf_search_unwind_table: e->fde_offset = 358, segbase = 10010878, debug_frame_base = 0, fde_addr = 10010bd0
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fe4)
 >_Uppc32_resume: (cursor=0x40ffe6cc)
FAILURE: b() returned from raise_exception()!!
FAILURE: test failed: result=0 got_here=1 nerrors=1
FAIL Gtest-exc (exit status: 255)

FAIL: Ltest-exc
===============

 >unw_init_local_common: (cursor=0x40ffe6cc)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000ce0)
 >_ULppc32_dwarf_search_unwind_table: e->fde_offset = 2a4, segbase = 1000ccf0, debug_frame_base = 0, fde_addr = 1000cf94
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000ef8)
 >_ULppc32_dwarf_search_unwind_table: e->fde_offset = 2ec, segbase = 1000ccf0, debug_frame_base = 0, fde_addr = 1000cfdc
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010001008)
 >_ULppc32_dwarf_search_unwind_table: e->fde_offset = 310, segbase = 1000ccf0, debug_frame_base = 0, fde_addr = 1000d000
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_dwarf_search_unwind_table: e->fde_offset = 310, segbase = 1000ccf0, debug_frame_base = 0, fde_addr = 1000d000
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_step: (cursor=0x40ffe6cc, ip=0x0000000010000fa4)
 >_ULppc32_resume: (cursor=0x40ffe6cc)
FAILURE: b() returned from raise_exception()!!
FAILURE: test failed: result=0 got_here=1 nerrors=1
FAIL Ltest-exc (exit status: 255)

[bertwesarg]

can you check if reverting my mi_mmap patch (#486) fixes this? In 32-bit big- or littel-endian, there exists an mmap2 syscall which accepts 64-bit arguements. I just recently learned about it, but could not figure out yet, if this really affects libunwind. The initiator was this post:

https://gcc.gnu.org/pipermail/gcc-patches/2023-June/622386.html

thanks

[bregma]

The failures appear to be in detecting the signal trampoline and in the unwind_resume code. I sincerely doubt it's a problem with mi_mmap().

That could be causing the issue on PPC64, I haven't looks into the large number of failures on those targets. I'll consider the possibility when that time comes.