On Sat, Dec 07, 2019 at 05:52:01PM +0300, Andrey Savchenko wrote:
> Hi all!
> 
> On Fri,  6 Dec 2019 18:36:55 +0300 Aleksei Nikiforov wrote:
> > This change should fix pointer arithmetic issues for e2k.
> 
> This commit message is misleading. Pointer arithmetic is broken on
> all architectures, but by pure chance in happens to work correctly
> when gcc is used. On e2k the lcc compiler is used, so problem
> manifests itself. But this doesn't change the fact that pointer
> arithmetic was flawed on all architectures.

Strictly speaking, it's not "broken", it's "UB", but I agree
it is not specific to e2k.

My original commit message for this change was the following:

"Fix UB in pointer arithmetic

Commit 6d5e6a689d07de8feef2cbecb24bc42d5994861b aka 0.5.15lorg2-alt70~9
among other changes introduced UB in pointer arithmetic by casting raw
pointers to specific types."


-- 
ldv