RFA: fix bugs in likely_spilled_retval_1 / likely_spilled_retval_p (Was: Re: RFA: Fix rtl-optimization/22258)

[Eric Botcazou]

>    Bootstrapped and tested on i686-pc-linux-gnu with no new failures and no
> new passes.

Great.  Can either of you install the fix on mainline?  It would also be nice 
to put a couple of ??? notes on the 4.1 and 4.2 branches (unless you can 
exhibit a regression that would be fixed by this patch, in which case we 
could consider putting it on the 4.2 branch too).

>    I noticed a a 16-byte increase in the size of the binaries. Before the
> patch:
>
>    text    data     bss     dec     hex filename
> 6501368   17600  567412 7086380  6c212c cc1
> 7518840   17600  577108 8113548  7bcd8c cc1plus
> 6608104   17600  569044 7194748  6dc87c cc1obj
> 6580620   23784  569416 7173820  6d76bc f951
>
>    After the patch:
>    text    data     bss     dec     hex filename
> 6501384   17600  567412 7086396  6c213c clean-gcc-i686/gcc/cc1
> 7518856   17600  577108 8113564  7bcd9c clean-gcc-i686/gcc/cc1plus
> 6608120   17600  569044 7194764  6dc88c clean-gcc-i686/gcc/cc1obj
> 6580636   23784  569416 7173836  6d76cc clean-gcc-i686/gcc/f951
>
>    I think this is caused by the extra line of code added by the patch and
> function alignment rather than by a missed optimization.

Looks sensible.  Thanks for conducting this comparison.

-- 
Eric Botcazou