> Ah. All validity issues aside, then "Dereferencing unaligned
> pointers yields a compile-time error" or "pointers with unknown
> alignment" would be much less cryptic: the "Dereferencing -1"
> just sounds like *(char *) -1 or a cut
I put a more explanatory comment in there.
I'm not exactly su
> This is unnecessary with my just applied series of libgcc patches:
> it's the default for all *-*-elf targets.
Moving targets! Fixed.
> This is not only unnecessary, as Joseph already noted, but doesn't
> work for quite some time since fp-bit.c has been moved to libgcc.
Fixed.
> > - unsigned int __max_iter = 100;
> > + unsigned int __max_iter = 65536U;
>
> Doesn't that need to be 65535U for your purpose?
Yup. The other three did.
> Should have the runtime license exception.
Added.
> This should not be needed; just using t-fdpbit should suffice for
>
> Wouldn't some tools behave better if the asm files had ELF
> decorations on the functions?
If you mean .type and .size, I added those.
Nobody has asked the obvious question: why does libssp use "*(int *)(-1)
= 0;" in the first place?
> Ah. All validity issues aside, then "Dereferencing unaligned
> pointers yields a compile-time error" or "pointers with unknown
> alignment" would be much less cryptic: the "Dereferencing -1"
> just sounds like *(char *) -1 or a cut
Good point.
On Wed, 2 Nov 2011, DJ Delorie wrote:
> > > Index: configure.ac
> > > +# Dereferencing -1 is a compile-time error
> >
> > This (those lines) look a little cryptic (and lack punctuation ;)
> > Wild improvement guess: "Too small 'int'?".
>
> No, the compiler specifically tests for unaligned acces
> deduce that this path is unreachable is to generate an abort and output an
> informative notice with inform ().
Hmmm... I'll see if I can catch it early enough to do something more
meaningful, then.
> (I don't see what actually generates the error, since there are only two
Actually, it's in
DJ Delorie writes:
> Index: libgcc/config.host
> ===
> --- libgcc/config.host(revision 180758)
> +++ libgcc/config.host(working copy)
> @@ -683,12 +683,16 @@ rs6000-ibm-aix5.1.* | powerpc-ibm-aix5.1
> tmake_file
On Wed, 2 Nov 2011, DJ Delorie wrote:
> > > Index: configure.ac
> > > +# Dereferencing -1 is a compile-time error
> >
> > This (those lines) look a little cryptic (and lack punctuation ;)
> > Wild improvement guess: "Too small 'int'?".
>
> No, the compiler specifically tests for unaligned ac
On Wed, 2 Nov 2011, DJ Delorie wrote:
> - unsigned int __max_iter = 100;
> + unsigned int __max_iter = 65536U;
Doesn't that need to be 65535U for your purpose?
> Index: libgcc/config/rl78/lib2shift.c
> ===
> --- libgcc
> > Index: configure.ac
> > +# Dereferencing -1 is a compile-time error
>
> This (those lines) look a little cryptic (and lack punctuation ;)
> Wild improvement guess: "Too small 'int'?".
No, the compiler specifically tests for unaligned accesses and gives a
compile-time error, at Renesas's
On Wed, 2 Nov 2011, DJ Delorie wrote:
> The Renesas RL78 is a new low-power 8/16 bit microcontroller, with an
> architecture much like the original Z80.
Just some random spottings.
> Index: configure.ac
> +# Dereferencing -1 is a compile-time error
This (those lines) look a little cryptic (a
13 matches
Mail list logo
