On Wed, Jan 28, 2026 at 11:09 AM Ihor Solodrai <[email protected]> wrote: > > On 1/28/26 10:49 AM, Alexei Starovoitov wrote: > > On Wed, Jan 28, 2026 at 9:34 AM Ihor Solodrai <[email protected]> > > wrote: > >> > >> On 1/28/26 7:35 AM, Chris Mason wrote: > >>> > >>> > >>> On 1/28/26 10:26 AM, [email protected] wrote: > >>>>> diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c > >>>>> index 5a075e06cf..070ba80e39 100644 > >>>>> --- a/arch/x86/net/bpf_jit_comp.c > >>>>> +++ b/arch/x86/net/bpf_jit_comp.c > >>>>> @@ -4112,3 +4112,8 @@ bool bpf_jit_supports_timed_may_goto(void) > >>>>> { > >>>>> return true; > >>>>> } > >>>>> + > >>>>> +bool bpf_jit_supports_fsession(void) > >>>> ^^^^^^^^^^^^^^^^^^^^^^^^^^ > >>>> This is the actual function name in the code. > >>> > >>> Ihor, I think the script parsing review-inline.txt chopped off the > >>> part of the review where AI was complaining about the commit message? > >> > >> This is the email body pre-processing in KPD, yes. > >> > >> At some point we decided to remove the commit message before sending > >> an email, but now that AI reviews the messages too, I think we should > >> just send the generated review-inline.txt as is. > >> > >> Alexei, wdyt? > > > > I think KPD is only supposed to trim the header until 'diff ..' line. > > In this case there is no 'diff', so I'm not sure why it trimmed so much. > > There is. Here is the PR comment with the review: > https://github.com/kernel-patches/bpf/pull/10868#issuecomment-3811930805 > > KPD trimmed everything before the first "> diff" occurrence, as expected. > > I'll try to fix this up in KPD. We should probably search for the first > line starting with "> " (a quote start) to trim the header correctly. > > Alternatively, AI can be prompted to avoid generating the header in > review-inline.txt, but that's probably less reliable. And maybe it is > useful for local runs, idk.
I see. let's then strip the first 4 lines ? (commit, author, subj) ? tbh this one was unusual. If it repeats then yeah '> diff' is a wrong marker.
