Bernhard Reutner-Fischer wrote:
>If i'm not mistaken, the s++ is fully redundant here although it seems
>that my gcc-11 does not optimize it like if we manually spell it out
>like in the attached?
I don't think it's possible to avoid two increments. When a backslash
is detected we want to move past it and the following character.
Consider:
strchr_backslash("\\/abc/def", '/');
The original code and Denys' rewrite return a pointer to "/def"; your
suggestion returns a pointer to "/abc/def".
>Must be marked noinline to avoid inlining it twice it seems?!
>If that double inlining really happens then the implied code growth
>even for -Os is a different bug..
Hmm, yes, that's rather an odd thing for the compiler to do.
Ron
_______________________________________________
busybox mailing list
[email protected]
http://lists.busybox.net/mailman/listinfo/busybox
