Support multiple --skip-after
#231
Conversation
perf is stopping short on some of my backtraces (I think it's related to complex C++ code in the stack), which fragments the relevant functions in the flamegraph. `--skip-after` works great to avoid this, but I want to do it for multiple functions.
bsilver8192
force-pushed
the
multiple-skip-after
branch
from
84744d9
to
2fb9aa4
Compare
Codecov Report
@@ Coverage Diff @@
## master #231 +/- ##
==========================================
- Coverage 87.01% 86.85% -0.16%
==========================================
Files 17 17
Lines 2379 2396 +17
==========================================
+ Hits 2070 2081 +11
- Misses 309 315 +6
Continue to review full report at Codecov.
|
Done. I went with adding a new one to make sure all the branches are covered (I assume that's what codecov is saying? I haven't dug into what numbers it's reporting). |
src/collapse/perf.rs
Outdated
| // without `skip_after` some collapsed lines would look like: | ||
| // go;[unknown];x_cgo_notify_runtime_init_done;runtime.main;main.init;... | ||
| if line.contains("main.init") { | ||
| assert!(line.contains("main.init;")); // we removed the frames above "main.init" |
There was a problem hiding this comment.
Actually, reading this test again. I think this should be line.starts_with, not line.contains. Otherwise it doesn't seem to test much?
There was a problem hiding this comment.
Good point. Fixed them both, and confirmed that the tests fail if I remove things from skip_after.
This follows jonhoo/inferno#231 Now we also support multiple --skip-after arguments. Update docu to reflect this. A semver bump will be required because FlamegraphOptions is public.
This follows jonhoo/inferno#231 Now we also support multiple --skip-after arguments. Update docu to reflect this. A semver bump will be required because FlamegraphOptions is public.
This follows jonhoo/inferno#231 Now we also support multiple --skip-after arguments. Update docu to reflect this. A semver bump will be required because FlamegraphOptions is public.
perf is stopping short on some of my backtraces (I think it's related to
complex C++ code in the stack), which fragments the relevant functions
in the flamegraph.
--skip-afterworks great to avoid this, but I wantto do it for multiple functions.