Allow Linux process command names longer than 15 chars

[lewisdoesstuff]

Currently, ProcessSample data on Linux hosts has a 15 character limit on the commandName value.
This is a limitation of the kernel - the comm value read from /proc/<pid>/stat is limited to TASK_COMM_LEN (16, including null byte).

This isn't the end of the world when collecting all processes, but causes issues when trying to limit the collected samples via include_matching_metrics. You can't currently match some-long-process-name-1 without using a wildcard to cover everything after the 15th character. If some-long-process-name-2 is also running, both of these will be sampled.

I've updated the pw.Command() method (snapshot_linux.go) to handle this.
If the process name (from /proc/<pid>/stat) is less than 15 chars, then we return that.
If it's 15, it could be truncated, so we use pw.process.Exe() to read the link at /proc/<pid>/exe, returning the base name.

This requires read permissions on the linked executable, so if the agent isn't running as root, the (possibly truncated) value is returned.

Comparison in events in ProcessSample - Top: patched, bottom: 1.30.0

[rajrohanyadav]

Hi @lewisdoesstuff
Thanks for your contribution. We will review and evaluate this PR soon for backwards compatibility and any potential breaking changes. We will try to merge this asap.
Thanks.

[rahulreddy15]

Hi @lewisdoesstuff
Thanks for your contribution. We are not going ahead with merging this PR because it could cause alerts that customers have setup to break. To safely release this, we would want to put enabling command names longer than 15 characters behind a feature flag as an infrastructure agent config option.
We have planned to prioritize this task and will provide an update here once the feature is part of GA.