On Thu, Sep 16, 2021 at 09:50:35PM +0200, Francesco Dolcini wrote:
> On Thu, Sep 16, 2021 at 03:42:20PM -0400, Tom Rini wrote:
> > On Thu, Sep 16, 2021 at 09:38:19PM +0200, Francesco Dolcini wrote:
> > > On Thu, Sep 16, 2021 at 01:30:21PM -0400, Tom Rini wrote:
> > > >  		*(.literal .text)					\
> > > > +		*(.literal .text_version_string)			\
> > > 
> > > Isn't ".litteral" a duplication? Even if I'm pretty sure it will not cause any difference
> > > in the generated binary I would remove it.
> > 
> > Honestly, I don't know xtensa.  We also don't have qemu support for it
> > currently, and I'm a bit worried in general about the state of the
> > platform (it's on my TODO list to poke some people now).  So, I hesitate
> > to make any change that's not basically copy/paste of the existing
> > lines.
> 
> I have no experience on xtensa either, but this is just the section
> name, and you are just telling to put ".literal" there 2 times.

I don't know.  Looking at the resulting linker script (and with your
suggestion):
  .text (((0x00002000 - 0x00002000) + ((128 << 20))) - 0x00040000) : AT(((LOADADDR(.DoubleExceptionVector.text) + SIZEOF(.DoubleExceptionVector.text) + 4 -1)) & ~(4 -1)) { _text_start = ABSOLUTE(.); *(.literal .text) *(.text_version_string) *(.literal.* .text.* .stub) *(.gnu.warning .gnu.linkonce.literal.*) *(.gnu.linkonce.t.*.literal .gnu.linkonce.t.*)
  *(.fini.literal) *(.fini) *(.gnu.version) _text_end = ABSOLUTE(.); }

So there's "literal" scattered everywhere.  If it doesn't matter, it
reads more consistently to me to toss literal in one more time.  But I
emailed around and we'll see if xtensa stays around, it's also our
oldest toolchain and virtually untouched since submission :(

-- 
Tom