New riscv CI and try builder for binutils, some ld failures

New riscv CI and try builder for binutils, some ld failures

[Mark Wielaard]

Hi,

There is a new binutils riscv builder using a StarFive VisionFive 2
board running ubuntu server 23.10:
https://builder.sourceware.org/buildbot/#/builders/binutils-ubuntu-riscv

There is also a try builder:
https://builder.sourceware.org/buildbot/#/builders/binutils-try-ubuntu-riscv

The gas and binutils testsuites are zero fail.
But there are four ld failures.

FAIL: Handle no DWARF information

I have a patch for that one.
[PATCH] bfd: riscv_maybe_function_sym check _bfd_elf_is_local_label_name

FAIL: Run pr21964-4

There seems to be some magic going on where a symbol starting with
__start should be made private even when declared extern. I don't
fully follow what is going on here.

FAIL: Run pr18841 with libpr18841c.so
FAIL: Run pr18841 with libpr18841cn.so (-z now)

Which seem to be caused by using the wrong kind of relocation when
the symbol is an ifunc. The testcase core dumps.

Cheers,

Mark

[PATCH] bfd: riscv_maybe_function_sym check _bfd_elf_is_local_label_name

[Mark Wielaard]

This fixes the ld "Handle no DWARF information" testcase. Which
currently fails on riscv because a local label name is assumed
to be the current function name.

bfd/ChangeLog:

	* elfnn-riscv.c (riscv_maybe_function_sym): Also check
	_bfd_elf_is_local_label_name.
---
 bfd/elfnn-riscv.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/bfd/elfnn-riscv.c b/bfd/elfnn-riscv.c
index 509d61e5017..15c7fa3f89e 100644
--- a/bfd/elfnn-riscv.c
+++ b/bfd/elfnn-riscv.c
@@ -5452,7 +5452,8 @@ riscv_maybe_function_sym (const asymbol *sym,
 			  bfd_vma *code_off)
 {
   if (sym->flags & BSF_LOCAL
-      && riscv_elf_is_mapping_symbols (sym->name))
+      && (riscv_elf_is_mapping_symbols (sym->name)
+	  || _bfd_elf_is_local_label_name (sec->owner, sym->name)))
     return 0;
 
   return _bfd_elf_maybe_function_sym (sym, sec, code_off);
-- 
2.39.3

Re: [PATCH] bfd: riscv_maybe_function_sym check _bfd_elf_is_local_label_name

[Mark Wielaard]

On Thu, 2024-01-04 at 01:31 +0100, Mark Wielaard wrote:
> This fixes the ld "Handle no DWARF information" testcase. Which
> currently fails on riscv because a local label name is assumed
> to be the current function name.
> 
> bfd/ChangeLog:
> 
> 	* elfnn-riscv.c (riscv_maybe_function_sym): Also check
> 	_bfd_elf_is_local_label_name.

Ping. OK to push?

> ---
>  bfd/elfnn-riscv.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/bfd/elfnn-riscv.c b/bfd/elfnn-riscv.c
> index 509d61e5017..15c7fa3f89e 100644
> --- a/bfd/elfnn-riscv.c
> +++ b/bfd/elfnn-riscv.c
> @@ -5452,7 +5452,8 @@ riscv_maybe_function_sym (const asymbol *sym,
>  			  bfd_vma *code_off)
>  {
>    if (sym->flags & BSF_LOCAL
> -      && riscv_elf_is_mapping_symbols (sym->name))
> +      && (riscv_elf_is_mapping_symbols (sym->name)
> +	  || _bfd_elf_is_local_label_name (sec->owner, sym->name)))
>      return 0;
>  
>    return _bfd_elf_maybe_function_sym (sym, sec, code_off);

Re: [PATCH] bfd: riscv_maybe_function_sym check _bfd_elf_is_local_label_name

[Nick Clifton]

Hi Mark,

> On Thu, 2024-01-04 at 01:31 +0100, Mark Wielaard wrote:
>> This fixes the ld "Handle no DWARF information" testcase. Which
>> currently fails on riscv because a local label name is assumed
>> to be the current function name.
>>
>> bfd/ChangeLog:
>>
>> 	* elfnn-riscv.c (riscv_maybe_function_sym): Also check
>> 	_bfd_elf_is_local_label_name.
> 
> Ping. OK to push?

Approved - please apply.

Cheers
   Nick