ASSERT synch.c:261 on client.destructor test on ubuntu22 32-bit

[derekbruening]

I tried to reproduce the synch.c:261 assert on the drcachesim tests on GA CI ubuntu22 32-bit seen in PR #6534 on a local machine and failed. But on running the whole suite there I did hit the assert in the client.destructor test.

Running under gdb, first we have SIGSEGV in privlib constructor:

Program received signal SIGSEGV, Segmentation fault.
0xf724fbed in ?? ()
(gdb) x/3i 0xf724fbed 
   0xf724fbed:	mov    %gs:0x4fe,%al
   0xf724fbf3:	movzbl %al,%eax
(gdb) bt
#0  0xf724fbed in ?? ()
#1  0xf72d649d in write ()
#2  0xf733dd65 in _IO_file_write ()
#3  0xf733d5a6 in ?? ()
#4  0xf733df15 in _IO_file_xsputn ()
#5  0xf723f904 in fwrite ()
#6  0xf78c5b49 in __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> >::xsputn(char const*, int) ()
#7  0xf78f2df2 in std::basic_ostream<char, std::char_traits<char> >& std::__ostream_insert<char, std::char_traits<char> >(std::basic_ostream<char, std::char_traits<char> >&, char const*, int) ()
#8  0xf78f3192 in std::basic_ostream<char, std::char_traits<char> >& std::operator<< <std::char_traits<char> >(std::basic_ostream<char, std::char_traits<char> >&, char const*) ()
#9  0xf7fba274 in hasdtr::hasdtr (this=0xf7fbd00f) at /home/derek/dr/git/src/suite/tests/client-interface/destructor.dll.cpp:46
#10 0xf7fba1ff in __static_initialization_and_destruction_0 (__initialize_p=1, __priority=65535) at /home/derek/dr/git/src/suite/tests/client-interface/destructor.dll.cpp:54
#11 0xf7fba247 in _GLOBAL__sub_I_destructor.dll.cpp(void) () at /home/derek/dr/git/src/suite/tests/client-interface/destructor.dll.cpp:60
#12 0xf7c8370a in privload_call_lib_func (func=0xf7fba228 <_GLOBAL__sub_I_destructor.dll.cpp(void)>) at /home/derek/dr/git/src/core/unix/loader.c:1079
#13 0xf7c8231b in privload_call_entry (dcontext=0x4075fb40, privmod=0x407584f8, reason=1) at /home/derek/dr/git/src/core/unix/loader.c:629
#14 0xf7b89823 in privload_call_entry_if_not_yet (dcontext=0x4075fb40, privmod=0x407584f8, reason=1) at /home/derek/dr/git/src/core/loader_shared.c:121
#15 0xf7b8b832 in privload_load_finalize (dcontext=0x4075fb40, privmod=0x407584f8) at /home/derek/dr/git/src/core/loader_shared.c:829
#16 0xf7b89baa in loader_init_epilogue (dcontext=0x4075fb40) at /home/derek/dr/git/src/core/loader_shared.c:217
#17 0xf7a3052f in dynamorio_app_init_part_two_finalize () at /home/derek/dr/git/src/core/dynamo.c:675
#18 0xf7c85b98 in privload_early_inject (sp=0xffffd4e0, old_libdr_base=0x0, old_libdr_size=0) at /home/derek/dr/git/src/core/unix/loader.c:2259
#19 0xf7c39a39 in reloaded_xfer () at /home/derek/dr/git/src/core/arch/x86/x86.asm:1187
#+end_example
Then we have a 2nd SIGSEGV processing the first:
#+begin_example
(gdb) c
Continuing.

Program received signal SIGSEGV, Segmentation fault.
0xf7b554c1 in check_in_last_thread_vm_area (dcontext=0x4075fb40, pc=0x4fe <error: Cannot access memory at address 0x4fe>) at /home/derek/dr/git/src/core/vmareas.c:8398
8398	        if (is_readable_without_exception((app_pc)&shared_data->last_area->end, 4) &&
(gdb) bt
#0  0xf7b554c1 in check_in_last_thread_vm_area (dcontext=0x4075fb40, pc=0x4fe <error: Cannot access memory at address 0x4fe>) at /home/derek/dr/git/src/core/vmareas.c:8398
#1  0xf7c75382 in main_signal_handler_C (xsp=0x407da3bc "R\232\303\367\v") at /home/derek/dr/git/src/core/unix/signal.c:5986
#2  0xf7c39ac1 in main_signal_handler () at /home/derek/dr/git/src/core/arch/x86/x86.asm:1460
(gdb) x/3i 0xf724fbed 
   0xf724fbed:	mov    %gs:0x4fe,%al
   0xf724fbf3:	movzbl %al,%eax
   0xf724fbf6:	movb   $0x1,%gs:0x4fe
(gdb) p shared_data
$1 = (thread_data_t *) 0x0

Hmm under gdb we don't get past this to hit the synch.c assert; from log file w/o debugger the assert is a little later, maybe while still trying to handle the 2nd SIGSEGV.

Fixing the check_in_last_thread_vm_area() by checking for shared_data==NULL, and wrapping privload_call_lib_func()'s call in TRY_EXCEPT: this test now doesn't die, but it fails and has a nice message:

240: Test command: /home/derek/dr/git/build_x86_dbg_tests/bin32/drrun "-s" "90" "-quiet" "-debug" "-killpg" "-stderr_mask" "0xC" "-dumpcore_mask" "0" "-code_api" "-c" "/home/derek/dr/git/build_x86_dbg_tests/suite/tests/bin/libclient.destructor.dll.so" "--" "/home/derek/dr/git/build_x86_dbg_tests/suite/tests/bin/client.destructor"
240: Test timeout computed to be: 1500
240: <Private library libclient.destructor.dll.so init/fini func 0xf7f44228 crashed>
240: all done
1/1 Test #240: code_api|client.destructor .......***Failed  Required regular expression not found. Regex=[^hasdtr::hasdtr
all done
hasdtr::~hasdtr
$
]  0.20 sec

This is yet more issues with private loading on recent glibc: xref #5437.