support new perf data

[Jason-Ltiger]

When parsing the perf.data file, the following error occurred. It seems that perfparse is not working properly to parse this file format：
QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-root'
QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-root'
kf.coreaddons: "Could not find plugin konsolepart"
MESA: error: Failed to query drm device.
libGL error: glx: failed to create dri3 screen
libGL error: failed to load driver: iris
libGL error: failed to open /dev/dri/card2: No such file or directory
libGL error: failed to load driver: iris
unhandled feature PerfHeader::SAMPLE_TIME 16
unhandled feature PerfHeader::CACHE 8
unhandled feature PerfHeader::BPF_PROG_INFO 4
unhandled feature PerfHeader::BPF_BTF 4
Linux version "6.1.119" detected. Switching to automatic buffering.
unhandled event type 79 PERF_RECORD_TIME_CONV
unhandled event type 69 PERF_RECORD_ID_INDEX
unhandled event type 73 PERF_RECORD_THREAD_MAP
unhandled event type 74 PERF_RECORD_CPU_MAP
unhandled event type 17 PERF_RECORD_KSYMBOL
unhandled event type 17 PERF_RECORD_KSYMBOL
unhandled event type 17 PERF_RECORD_KSYMBOL
unhandled event type 17 PERF_RECORD_KSYMBOL
unhandled event type 17 PERF_RECORD_KSYMBOL
unhandled event type 17 PERF_RECORD_KSYMBOL
unhandled event type 17 PERF_RECORD_KSYMBOL
unhandled event type 17 PERF_RECORD_KSYMBOL
unhandled event type 17 PERF_RECORD_KSYMBOL
unhandled event type 17 PERF_RECORD_KSYMBOL
unhandled event type 82 unknown type

[Jason-Ltiger]

When I am parsing the perf file recorded by ARM on an x86 machine
This is my Arm machine version：
nvidia@tegra-ubuntu:$ perf --version
perf version 6.1.119
nvidia@tegra-ubuntu:$ uname -a
Linux tegra-ubuntu 6.1.119-rt45-prod-rt-tegra #1 SMP PREEMPT_RT Wed May 14 21:57:54 UTC 2025 aarch64 aarch64 aarch64 GNU/Linux

[milianw]

can you submit one of such data files without compromising any internal information? how do you record the data file? what version of hotspot are you even using?

[GitMensch]

Are those just warnings or does anything chrash or stop doing the analysis or do you "just" see no data?
@milianw: should perfparser handle PERF_RECORD_KSYMBOL with its struct perf_record_ksymbol?

I think that perfparser may explicit throw away any event >= PERF_RECORD_MAX "non-ABI", which is 22 for kernel master - @milianw, what do you think of me adding a check to the default switch in PerfData::processEvents() to either ignore them silently or, as separate case (also with qWarning or even qInfo, noting something about "non-ABI event %d ignored."?

If I don't get a minimal perf.data example that PR wouldn't be covered by a testcase, though.

[milianw]

as evidenced by #736 we really should not ignore new stuff that simply got added since we last with the kernel.

to actually understand what the problem is here we'd need more data and information. which we haven't received from OP so far. I'll close this for now. Please reopen with more details if anyone can provide it.