Re: CVS commit: src/sys/arch

[Joerg Sonnenberger <joerg%bec.de@localhost>]

On Tue, Aug 08, 2023 at 04:01:19PM +1000, matthew green wrote:
> Joerg Sonnenberger writes:
> > On Thu, Aug 03, 2023 at 08:16:31AM +0000, matthew green wrote:
> > > Module Name:	src
> > > Committed By:	mrg
> > > Date:		Thu Aug  3 08:16:31 UTC 2023
> > > 
> > > Modified Files:
> > > 	src/sys/arch/evbarm/gumstix: gumstix_machdep.c
> > > 	src/sys/arch/evbarm/ixm1200: ixm1200_machdep.c
> > > 	src/sys/arch/hpcarm/hpcarm: pxa2x0_hpc_machdep.c sa11x0_hpc_machdep.c
> > > 	src/sys/arch/hppa/stand: Makefile.buildboot
> > > 	src/sys/arch/m68k/m68k: regdump.c
> > > 	src/sys/arch/macppc/macppc: cpu.c
> > > 
> > > Log Message:
> > > ignore "-Warray-bounds" for various low level platform code that knows
> > > how something is setup but technically is undefined behaviour.  the
> > > most common here is "extern int end;" and then using offsets of "&end"
> > > that are outside the bounds of this 4-byte integer.
> > > 
> > > these uses are almost certainly all OK in reality.
> >
> > Are you sure that GCC's optimizer is not going to break any of those
> > assumptions? We had to go through quite some trouble in crtbegin.c for
> > similar patterns?
> 
> nope, infact i'm not.  i'll revert these.
> 
> do you have good ways to fix these from that crt issue?

Not really, I just remember that we needed a few iterations. The last
round was especially mysterious as weak references suddenly couldn't
alias extern symbol.

Joerg