On 01/08/2022 13:45, Arnaldo Carvalho de Melo wrote:
> Em Sun, Jul 31, 2022 at 06:38:26PM -0700, Andres Freund escreveu:
>> binutils changed the signature of init_disassemble_info(), which now causes
>> compilation failures for tools/{perf,bpf} on e.g. debian unstable. Relevant
>> binutils commit:
>> https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=60a3da00bd5407f07
>>
>> I first fixed this without introducing the compat header, as suggested by
>> Quentin, but I thought the amount of repeated boilerplate was a bit too
>> much. So instead I introduced a compat header to wrap the API changes. Even
>> tools/bpf/bpftool/jit_disasm.c, which needs its own callbacks for json, imo
>> looks nicer this way.
>>
>> I'm not regular contributor, so it very well might be my procedures are a
>> bit off...
>>
>> I am not sure I added the right [number of] people to CC?
>
> I think its ok
>
>> WRT the feature test: Not sure what the point of the -DPACKAGE='"perf"' is,
>
> I think its related to libbfd, and it comes from a long time ago, trying
> to find the cset adding that...
>
>> nor why tools/perf/Makefile.config sets some LDFLAGS/CFLAGS that are also
>> in feature/Makefile and why -ldl isn't needed in the other places. But...
>>
>> V2:
>> - split patches further, so that tools/bpf and tools/perf part are entirely
>> separate
>
> Cool, thanks, I'll process the first 4 patches, then at some point the
> bpftool bits can be merged, alternatively I can process those as well if
> the bpftool maintainers are ok with it.
>
> I'll just wait a bit to see if Jiri and others have something to say.
>
> - Arnaldo
Thanks for this work! For the series:
Acked-by: Quentin Monnet <quentin@xxxxxxxxxxxxx>
For what it's worth, it would make sense to me that these patches remain
together (so, through Arnaldo's tree), given that both the perf and
bpftool parts depend on dis-asm-compat.h being available.
Quentin
