[PATCH] gas: Document @ and % as interchangeable for type argument in .section

[Will Hawkins]

On Mon, May 18, 2026 at 9:17 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> On 18.05.2026 15:00, Will Hawkins wrote:
> > On Mon, May 18, 2026 at 1:57 AM Jan Beulich <jbeulich@suse.com> wrote:
> >> On 17.05.2026 13:35, Will Hawkins wrote:
> >>> On Sunday, May 17, 2026, Will Hawkins <hawkinsw@obs.cr> wrote:
> >>>> On Sunday, May 17, 2026, Jan Beulich <jbeulich@suse.com> wrote:
> >>>>> On 16.05.2026 02:38, Will Hawkins wrote:
> >>>>>> In the .section Assembler Directive for the ELF format, both the % and
> >>>>>> the @ can be used when writing the type argument.
> >>>>>
> >>>>> That's not exactly true (the two aren't fully interchangeable), and hence
> >>>>> ...
> >>>>
> >>>>
> >>>>
> >>>> In obj-elf.c, I saw
> >>>>
> >>>> else if (c == '@' || c == '%')
> >>>>
> >>>> which I must have misinterpreted?
> >>>
> >>> I guess that they are only treated equivalently in code on platforms where
> >>> the @ and % do not have implications for the code in earlier processing
> >>> before reaching obj_elf_section.
> >>
> >> Correct, comments are being removed much earlier. (If they were fully
> >
> > I am glad that my reading of the code was not wildly inaccurate: _If_
> > there were no preprocessor implications for those characters, then
> > they would be equivalent.
> >
> >> equivalent, in that testcase of yours you also wouldn't have had to
> >> switch from @ to %.)
> >
> > Yes, I agree.
> >
> > Before I propose an alternate, I figured it might be worth describing
> > why I wanted to make the change in the first place:
> >
> > My local system is an x86-64. When I started developing a test case
> > for objdump and I was running the testsuite, the test in objdump.exp
> > based on the code in dw5-op.S passed. I could read the documentation
> > and see that @progbits should be the syntax.
> >
> > I interpreted the existing note to imply mutually exclusive options
> > for @ vs %. And yet the test was passing. I used as to compile it --
> > just to be sure. Once I dug into the code and saw the aforementioned
> > if statement, I realized _why_ the test author used %: because % is
> > not a comment indicator on x86-64, using the % means that one version
> > of the test code works for both x86-64 and ARM platforms.
> >
> > I proposed the change so that others would not have to search as much
> > as I did. With that as background, would you be amenable to rewriting
> >
> > Note on targets where the @ character is the start of a comment (eg
> > ARM) then another character is used instead. For example the ARM port
> > uses the % character.
> >
> > to
> >
> > Note - certain targets treat @ as the start of a comment (e.g., ARM).
> > On such platforms, % is used instead. On platforms where neither @ nor
> > % indicate the start of a comment, @ and % may be used
> > interchangeably.
>
> This reads okay to me.

Thank you! I will send a v2!
Will


>
> Jan