https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90586
--- Comment #2 from Yibiao Yang <yangyibiao at nju dot edu.cn> ---
(In reply to Richard Biener from comment #1)
> The reason is that somehow the only stmt with line 9 prevailing is one
> not originally having it:
>
> <bb 6> :
> [t.c:11:14] D.1918 = 0;
> [t.c:11:14] // predicted unlikely by early return (on trees) predictor.
> [t.c:9:11] g = {CLOBBER};
> goto <bb 8>; [INV]
>
> <bb 7> :
> [t.c:14:10] D.1918 = 0;
>
> <bb 8> :
> <L8>:
> return D.1918;
>
> so the return stmt will end up inheriting it if we are "lucky". try-finally
> lowering assigns the location, possibly simply taking the location of the
> last stmt from the "throw".
Thanks for the explanation. Thus, I was wondering that this bug should be
marked as "invalid"
