thread-instruction-count: tidelta / ticount don't exist in tracing file

Environment 1 puppeteer-core: 2.0.0 docker image: node:12-slim os version: macOS 10.14.6 node.js version: 12 chrome version: 77/78

Environment 2 puppeteer-core: 2.0.0 virtualisation: virtualbox 6 os version: Ubuntu 18.04 node.js version: 12 chrome version: 77

What steps will reproduce the problem? According to this message. Puppeteer provides RTI data in trace file when flags: --enable-thread-instruction-count and --no-sandbox are used. Information should be placed in tidelta and ticount properties.

Case:

const puppeteer = require('puppeteer')

;(async () => {
  const browser = await puppeteer.launch({
    args: ['--no-sandbox', '--enable-thread-instruction-count'],
  })
  const page = await browser.newPage()

  await page.tracing.start({ path: './trace.json' })
  await page.goto('https://www.example.com/')
  await page.tracing.stop()

  await browser.close()
})()

Dockerfile Chrome-Detect&Downloader

What is the expected result? tidelta exist in trace.json and parameter not empty.

What happens instead? tidelta doesn’t exist in trace.json.

What a goal to use this feature? I want to use RTI data for Web Perf Testing in tools like size-limit, estimo, etc.

PS I tried this case with docker image node:10-slim, chrome 77 and chrome 78. But tidelta always didn’t exist in trace file.

[mitermayer]

I just noticed that I was comparing two different chrome versions:

  "os-version": "5.5.13-arch2-1",
  "product-version": "HeadlessChrome/80.0.3987.163",
  "v8-version": "8.0.426.30"

  "os-version": "5.5.13-arch2-1",
  "product-version": "HeadlessChrome/84.0.4147.13",
  "v8-version": "8.4.371.7"

Will update chrome and test again

[MurzNN]

When I manually launch chrome and enable tracing, I see in output

$ ./chrome --no-sandbox --enable-thread-instruction-count
[329820:329820:0609/065603.823454:ERROR:sandbox_linux.cc(369)] InitializeSandbox() called with multiple threads in process gpu-process.
[329820:329820:0609/065605.305150:ERROR:buffer_manager.cc(488)] [.DisplayCompositor]GL ERROR :GL_INVALID_OPERATION : glBufferData: <- error from previous GL command
tracing_service_impl.cc: Configured tracing, #sources:2, duration:0 ms, #buffers:1, total buffer size:102400 KB, total sessions:1
[329975:329977:0609/065638.525405:ERROR:thread_instruction_count.cc(65)] perf_event_open failed, omitting instruction counters
[329975:329975:0609/065638.524778:ERROR:thread_instruction_count.cc(65)] perf_event_open failed, omitting instruction counters
[329822:329856:0609/065638.528999:ERROR:thread_instruction_count.cc(65)] perf_event_open failed, omitting instruction counters
[329975:329978:0609/065638.528997:ERROR:thread_instruction_count.cc(65)] perf_event_open failed, omitting instruction counters
[329822:329853:0609/065638.529197:ERROR:thread_instruction_count.cc(65)] perf_event_open failed, omitting instruction counters
[329820:329847:0609/065638.535137:ERROR:thread_instruction_count.cc(65)] perf_event_open failed, omitting instruction counters
[329894:329894:0609/065638.535123:ERROR:thread_instruction_count.cc(65)] perf_event_open failed, omitting instruction counters

Seems perf_event_open failed is source of problem?

And magic command “allow kernel profiling for non-privileged users”, that I find here - seems help:

echo 1 | sudo tee /proc/sys/kernel/perf_event_paranoid

After enabling it - I see ticount in json results of your test file!