not sure how to test this - it seems that even setting lazy_symbolize to numbers other than 1 or 0 don't cause any "noticeable" issues - maybe the errors get swallowed.

The other motivation for this change is so Meta doesn't have to supply a patch to hard code this value to 1 due to the size of some of the binaries that it traces against.

Should we also add this as a config option?

Should we also add this as a config option?

I started down that path but because this is more specific to the environment or binary target (so yeah maybe a case for specific uprobes), I thought it better as just an env var but I don't feel strongly.

I think "symbolication" is the more accepted term over "symbolization". Some quick searching backs this up, e.g. https://lldb.llvm.org/use/symbolication.html

I'm also biased because I don't like the American spelling of "symbolisation" ;)

Ah great. @ajor is starting a culture war...

I honestly don't feel strongly though BPFTRACE_LAZY_SYMBOLICATE is more letters AND the actual bcc field is "lazy_symbolize". That being said, in casual conversation I do tend to say "symbolication" so I yield to the wise recommendations of the bpftrace maintainers. But "analyse" is wrong.

Should we also add this as a config option?

I started down that path but because this is more specific to the environment or binary target (so yeah maybe a case for specific uprobes), I thought it better as just an env var but I don't feel strongly.

The scripts usually contain the binary target (as a part of uprobes), so IMHO it makes sense to be able to set this in the script via a config option.

Ok, changed it to 'BPFTRACE_LAZY_SYMBOLICATION' and added it to the config.