[PATCH] gas: copy st_size only if unset
Alan Modra
amodra@gmail.com
Sat Apr 9 01:12:32 GMT 2022
On Fri, Apr 08, 2022 at 04:01:13PM -0700, Fangrui Song wrote:
> On 2022-04-08, Fangrui Song wrote:
> > On 2022-04-08, Alan Modra wrote:
> > > Your patch regressed these tests. Please investigate.
> > >
> > > aarch64_be-linux-gnu_ilp32 +FAIL: ELF symbol size
> > > aarch64-elf +FAIL: ELF symbol size
> > > aarch64-linux +FAIL: ELF symbol size
> > > alpha-linux +FAIL: ELF symbol size
> > > alpha-netbsd +FAIL: ELF symbol size
> > > alpha-unknown-freebsd4.7 +FAIL: ELF symbol size
> > > hppa64-hp-hpux11.23 +FAIL: ELF symbol size
> > > hppa64-linux +XPASS: .reloc against undefined local symbol (PR 27228)
> > > hppa-linux +XPASS: .reloc against undefined local symbol (PR 27228)
> > > nds32be-elf +FAIL: readelf -wiaoRlL dw5
> > > nds32le-linux +FAIL: readelf -wiaoRlL dw5
> > > riscv32-elf +FAIL: ELF symbol size
> > > riscv64-linux +FAIL: ELF symbol size
> > >
> > > --
> > > Alan Modra
> > > Australia Development Lab, IBM
> >
> > Sorry for the breakage. a3a7f5e1586467b137b8dcdcd2f74f5efa9f3919 should fix aarch64/alpha/hppa/riscv.
> > I am puzzled by
> >
> > > nds32be-elf +FAIL: readelf -wiaoRlL dw5
> > > nds32le-linux +FAIL: readelf -wiaoRlL dw5
>
> OK, I think this is related to Nick's
> 19c26da69d68d5d863f37c06ad73ab6292d02ffa
> ("Add code to display the contents of .debug_loclists sections which contain offset entry tables.")
>
> There is a diagnostic which has moved from somewhere in the middle to
> the top:
>
> readelf: Warning: unable to apply unsupported type 208 to section .debug_loclists
>
> Adding #... as the first line will fix the test for nds32*, but I am
> unsure whether the test should be adjusted to unsupport nds32*.
You are exactly correct in your analysis. Or we are both wrong in the
same way, because I independently came to the same conclusion. :-)
The test shouldn't be adjusted. It's now showing a defect in readelf
support for nds32 that was accidentally hidden before Nick's change.
--
Alan Modra
Australia Development Lab, IBM
More information about the Binutils
mailing list