Tool assumes utf8

[beertje44]

First of all: this is just very cool and promising! Up until now I was experimenting with pyrospy (written in php and trying about the same as you do in go) and it was just no good at all: it ran for a couple of seconds and produced traces during that time frame but always bailed out after a short while. Basically it was useless 😄 Your tool seems very stable and well made 👍

Anyway before I put it to real use, I liked to gave it a test spin on my personal wordpress blog that runs on a modified bitnami-wordpress-nginx docker image. As it turned out bitnami does not include a utf8 locale by default: bitnami/minideb#151 This broke your tool. It looks like you assume utf8 and when the source isn't valid utf8, pyroscope bails out with a HTTP 402 error and logs something like this:

invalid utf8 string hex: 6d61727368616c6c696e672e707920ffffffff2e6d616b65

I managed to set the locale to en_US.UTF-8 but still sometimes still get these errors. Perhaps all that is needed is a check or rewrite to valid utf8 in your pipelines?

[beertje44]

After so more debugging it was the c in qcup you use as default in the request-info arg to phpspy. Guess too many cookies lead to a too large label in pyroscope and it cuts them off but apparently this then leads to invalid utf8.

Looks like a bug upstream to me, but for now I just dropped cookies all together and all is well now 😄 For whats its worth: occionally I see utf8 encoding errors with the generic Alloy Pyroscope profiling for docker containers.