[Bug tree-optimization/97164] [8/9/10/11 Regression] incorrect offset on structure member where type of that member has aligned attribute

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97164

--- Comment #8 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #7)
> With additional
> --- libgcc/config/libbid/bid_functions.h.jj	2020-01-14 20:02:48.619582332
> +0100
> +++ libgcc/config/libbid/bid_functions.h	2020-09-23 01:12:02.672546190 +0200
> @@ -81,7 +81,7 @@ ALIGN (16)
>  #define SQRT80 sqrtw
>  #endif
>  
> -     typedef ALIGN (16)
> +     typedef ALIGN (8)
>       struct {
>         UINT64 w[3];
>       } UINT192;
> it bootstrapped on x86_64-linux, with
> gcc.c-torture/execute/pr36093.c and gcc.c-torture/execute/pr43783.c tests
> that use the same things regressing (guess they'd need to be removed).
> Is libgcc always compiled with the new compiler rather than with system
> compiler though?  If not, that might be a blocker for building older gcc
> versions with the new one.

Yes, libgcc is always built with the built compiler since it's a target
library.
The exception might be for canadian cross but IIRC we do not support mixing
compiler versions there?