[PATCH 0/13] [NDS32] Update and support more target features for the nds32 port

[Nelson Chu]

Hi Nick,

Thank you for getting back to me :)

1. I have fixed the first build errors in the tenth patch
(0010-NDS32-Update-the-gp-related-relaxations). The modified patch
will be sent again in the previous related mail.

2. For the strip-13 test case, we set the unsupported relocation from
215 to 255 for NDS32. It will be fixed in the thirteenth patch
(0013-NDS32-Support-TLS-LE-IE-IEGP-DESC-GD-LD) since the number 215 is
used by the new TLS reloc `R_NDS32_TLS_DESC_FUNC'.

3. For the "readelf -r bintest", R_NDS32_DATA is used for some
specific link-time relaxations, but these relaxations are obsolete.
Therefore, we don't need to generate the R_NDS32_DATA so far. It will
be fixed in the first patch
(0001-NDS32-Remove-the-unsupported-target-features) and we don't need
to amend
the readelf.r.

Finally, I am very sorry for the inconvenience that combine the
changelogs with the patches. Therefore, I will sent these patches
again in the previous related mails..

Thank you again for everything you’ve done :)

Best regards,
Nelson Ching-Yu Chu
Software R&D, Andes Technology
Nick Clifton <nickc@redhat.com> 於 2018年9月18日 週二 下午9:32寫道：
>
> Hi Nelson,
>
>   Thank you very much for submitting this patch series.
>
>   I did encounter a couple of problems building the patched files however:
>
> bfd/elf32-nds32.c: In function 'nds32_elf_relax_section':
> bfd/elf32-nds32.c:12241:37: error: bitwise comparison always evaluates to false [-Werror=tautological-compare]
>        || (sec->flags & SEC_EXCLUDE) == 1
>                                      ^~
>   and:
>
> bfd/elf32-nds32.c: In function 'nds32_elf_relocate_section':
> bfd/elf32-nds32.c:5271:11: error: this statement may fall through [-Werror=implicit-fallthrough=]
>         if (strcmp (elf_sym_hashes (input_bfd)[symndx]->root.root.string,
>            ^
> bfd/elf32-nds32.c:5283:6: note: here
>       case R_NDS32_SDA19S0_RELA:
>       ^~~~
>
>   This is with gcc v 8.1.1 on a Fedora 28 box.  They should be very easy to fix.
>
>
>   I also see two regressions in the binutils testsuite with the patches applied:
>
>     Checking Binutils in: nds32le-elf ... GAS: 8  LD: 22  BIN: 7  done
>      BIN REGRESSION: binutils-all/strip-13
>      BIN REGRESSION: readelf -r bintest
>
>   The strip 13 regression is because:
>
>     succeeded with: <>, expected: <\A[^\n]*: unsupported relocation type 0x[0-9a-f]+\n[^\n]*: bad value\Z>
>     FAIL: binutils-all/strip-13
>
>   So you may need to patch binutils/testsuite/binutils-all/objcopy.exp and
>   choose a different number for the "unsupported" reloc when running the
>   strip-13 test for the NDS.  (See the code immediately following which does
>   something similar for the strip-14 test).
>
>   For the "readelf -r bintest" case, it looks like you will need to amend
>   binutils/testsuite/binutils-all/readelf.r.  There is already some NDS specific
>   code in there.  You will probably need to add some more.
>
>   (As a general guideline, please run all of the binutils testsuites before
>   submitting a patch, and check/fix any regressions).
>
>
>   One other thing, in the future, please could you provide the
>   changelog entries as plain text, rather than as part of the
>   diff.  The problem is that the changelogs change so frequntly
>   that the diffs almost never apply.  In fact when I was applying
>   the patch series, every single one of them failed because the
>   changelog frags would not apply.
>
> Cheers
>   Nick