In the past I've used the profiler on a Mac with Intel processor (with default features) and on a M1 Mac (with frame-pointer feature).
Everything worked fine.
I've tried to run the same program on Linux (default features) and got several problem:
if I didn't remove pthread from the blocklist ["libc", "libgcc", "pthread", "vdso"] it would not collect samples. Actually it would collect 2 samples in total, no matter how long I profiled, with only 2 frames belonging to the signal_handler and nothing else
removing pthread (or the whole blocklist) most of the time resulted in the program crashing, for the reason explained in the readme as well
when the program was not crashing, the resulting stack traces were incoherent. For example, the frame belonging to the main function sometimes appears twice in the same stack traces and in different positions
I've run this on the following system: Debian 4.19.249-2 x86_64 GNU/Linux
And the version of the pprof-rs crate I've used is: 0.10.0
In the past I've used the profiler on a Mac with Intel processor (with default features) and on a M1 Mac (with frame-pointer feature). Everything worked fine.
I've tried to run the same program on Linux (default features) and got several problem:
pthread
from the blocklist["libc", "libgcc", "pthread", "vdso"]
it would not collect samples. Actually it would collect 2 samples in total, no matter how long I profiled, with only 2 frames belonging to thesignal_handler
and nothing elsepthread
(or the whole blocklist) most of the time resulted in the program crashing, for the reason explained in the readme as wellmain
function sometimes appears twice in the same stack traces and in different positionsI've run this on the following system:
Debian 4.19.249-2 x86_64 GNU/Linux
And the version of thepprof-rs
crate I've used is:0.10.0