Re: [PATCH 28/28] Decouple inferior_ptid/inferior_thread(); dup ptids in thread list (PR/25412)

[Pedro Alves <palves@redhat.com>]

On 6/23/20 2:37 PM, Andrew Burgess wrote:

> This commit causes a regression for gdb.gdb/unittest.exp when GDB is
> configured with --enable-targets=all.
> 
> The failure is:
> 
>   gdb/thread.c:95: internal-error: thread_info* inferior_thread(): Assertion `current_thread_ != nullptr' failed.
> 
> Here's a partial backtrace:
> 
>   #8  0x00000000013e4d0f in internal_error (file=0x16c55c7 "../../src.dev-1/gdb/thread.c", line=95, 
>       fmt=0x16c54d9 "%s: Assertion `%s' failed.") at ../../src.dev-1/gdbsupport/errors.cc:55
>   #9  0x0000000000c34e63 in inferior_thread () at ../../src.dev-1/gdb/thread.c:95
>   #10 0x0000000000a73d62 in get_current_regcache () at ../../src.dev-1/gdb/regcache.c:391
>   #11 0x00000000009051c0 in current_options () at ../../src.dev-1/gdb/mep-tdep.c:873
>   #12 0x00000000009052eb in mep_register_name (gdbarch=0x33862a0, regnr=152) at ../../src.dev-1/gdb/mep-tdep.c:958
>   #13 0x00000000009054aa in mep_register_reggroup_p (gdbarch=0x33862a0, regnum=152, group=0x26334c0 <save_group>)
>       at ../../src.dev-1/gdb/mep-tdep.c:1029
>   #14 0x00000000007975e8 in gdbarch_register_reggroup_p (gdbarch=0x33862a0, regnum=152, reggroup=0x26334c0 <save_group>)
>       at ../../src.dev-1/gdb/gdbarch.c:3595
>   #15 0x0000000000a7355a in reg_buffer::save(gdb::function_view<register_status (int, unsigned char*)>) (
>       this=0x7fffffffad00, cooked_read=...) at ../../src.dev-1/gdb/regcache.c:248
>   #16 0x0000000000780b36 in readonly_detached_regcache::readonly_detached_regcache(gdbarch*, gdb::function_view<register_status (int, unsigned char*)>) (this=0x7fffffffad00, gdbarch=0x33862a0, cooked_read=...)
>       at ../../src.dev-1/gdb/regcache.h:455
>   #17 0x0000000000a73409 in readonly_detached_regcache::readonly_detached_regcache (this=0x7fffffffad00, src=...)
>       at ../../src.dev-1/gdb/regcache.c:213
>   #18 0x0000000000a773b5 in selftests::cooked_read_test (gdbarch=0x33862a0) at ../../src.dev-1/gdb/regcache.c:1690
>   #19 0x0000000000b48c29 in selftests::gdbarch_selftest::operator() (this=0x320fef0)
>       at ../../src.dev-1/gdb/selftest-arch.c:73
>   #20 0x00000000013fe5ac in selftests::run_tests (filter=0x0) at ../../src.dev-1/gdbsupport/selftest.cc:88
>   #21 0x00000000008e78ba in maintenance_selftest (args=0x0, from_tty=1) at ../../src.dev-1/gdb/maint.c:1044
>   #22 0x00000000005b0557 in do_const_cfunc (c=0x32004f0, args=0x0, from_tty=1)
>       at ../../src.dev-1/gdb/cli/cli-decode.c:95
>   #23 0x00000000005b3839 in cmd_func (cmd=0x32004f0, args=0x0, from_tty=1) at ../../src.dev-1/gdb/cli/cli-decode.c:2113
>   #24 0x0000000000c4953b in execute_command (p=0x2924895 "", from_tty=1) at ../../src.dev-1/gdb/top.c:655
>   #25 0x0000000000759c26 in command_handler (command=0x2924880 "maintenance selftest ")
>       at ../../src.dev-1/gdb/event-top.c:588
>   #26 0x000000000075a052 in command_line_handler (rl=...) at ../../src.dev-1/gdb/event-top.c:773
> 
> I suspect that the problem might be this line in regcache.c:cooked_read_test:
> 
>   /* Switch to the mock thread.  */
>   scoped_restore restore_inferior_ptid
>     = make_scoped_restore (&inferior_ptid, mock_ptid);
> 
> My suspicion from a quick read of your patch above is that we need to
> do more than just set inferior_ptid here now.

Indeed.  Thanks for reporting it.  

I fixed this in gdbarch-selftests.c, where this code is duplicated,
but missed regcache.c.  I'll fix it.

Thanks,
Pedro Alves