Home Original page

☠ Buildbot (Sourceware): binutils-gdb - failed test (failure) test (failure) (master)

Palmer Dabbelt palmer@dabbelt.com
Tue Apr 16 15:54:51 GMT 2024 
On Tue, 16 Apr 2024 08:45:00 PDT (-0700), Nick Clifton wrote:
> Hi Mark,
>
>> Groan, sorry, it seems to be a real regression on riscv (in binutils).
>> But because ld did see failures before buildbot seems to get confused
>> who to "blame"...
>
>> Caused by:
>>      Remove accidental commit of an experimental change
>
>
>> diff --git a/binutils/testsuite/binutils-all/pr25662.ld
>> b/binutils/testsuite/binutils-all/pr25662.ld
>> index 4951184f88ef..19ef1391f8d9 100644
>> --- a/binutils/testsuite/binutils-all/pr25662.ld
>> +++ b/binutils/testsuite/binutils-all/pr25662.ld
>> @@ -12,6 +12,4 @@ SECTIONS
>>     .text : { *(.text) } > ROM
>>
>>     .bss : { *(.bss) } > RAM
>> -
>> -  /DISCARD/ : { *(.*) }
>>   }
>>
>> So maybe this testcase accidentially succeeded on riscv, but was
>> supposed to fail?
>
> Doh - now I remember what I was trying to fix.  The Risc-V targets
> are creating a RISCV_ATTRIBUTES section which gets moved to a new
> address when the binary is objcopy'ed.  I was wondering if special
> sections like this were expected to survive the copy completely
> unchanged, and decided to try dropping them the link entirely.  But
> I failed to follow up on why the RISCV_ATTRIBUTES section moves
> and - as Alan pointed out - dropping special sections creates new
> problems for other targets.
>
> So I need to go back to the Risc-V code in the BFD backend and see
> if I can find out why this section is causing problems for objcopy.

I can't think of any reason the RISC-V attributes would specifically 
trip up on an objcopy that doesn't change anything else.  We've got some 
string-based attributes, but if nothing else changes they should stay 
the same too.  There's also some attribute merging code, but again that 
shouldn't change anything.

That said, I wouldn't be surprised if we have some bug floating around 
there somewhere.  The attributes aren't all that widely used (they sort 
of just store a bunch of stuff that doesn't have much meaning, we've 
gotten burned by backwards compatability there a few times).  I think 
tooling mostly ignores them these days.

LMK if you want Nelson or I to look, but happpy to have the help ;)

>
> Cheers
>    Nick
More information about the Binutils mailing list