Comment # 19 on bug 1224402 from Miroslav Beneš

(In reply to Martin Jambor from comment #13)
> GCC 13 (at least) does not automagically promote the array to const because
> of the -flive-patching=inline-clone option (which disables the pure-const
> pass).  It is even natural, we need to be ready for some unknown live-patch
> to possibly leak the address even when such a leak is not present in the
> code at hand.

Bah, right. I wonder if we should just get rid of the option eventually as
upstream did and somehow detect dangerous changes in our tooling. Adding Petr
for his awareness. Thanks a lot for the explanation.

(In reply to Stefan Dirsch from comment #18)
> Ok. BTW, there is no such warning when building against Tumbleweed, just
> against Leap 15.5 and 15.6. Check this
> 
> https://build.suse.de/project/monitor/Proprietary:X11:Drivers
> 
> nvidia-driver-G06
> nvidia-gfxG05
> nvidia-gfxG04 (this bug G04 = 390.157)
> 
> All these are affected by this objtool warning on Leap. mabye one should
> figure out what's different in objtool for Tumbleweed?

Upstream objtool reports the same warning on the object file so there is no
difference. TW does not use the GCC option mentioned above so the object code
is different.