Add a test for permission denied test case

[danielocfb]

This change adds a test checking expected behavior when access to a symbolization source file is denied.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (cdde707) 92.30% compared to head (3ca1660) 92.26%.

❗ Current head 3ca1660 differs from pull request most recent head 9163a49. Consider uploading reports for the commit 9163a49 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #448      +/-   ##
==========================================
- Coverage   92.30%   92.26%   -0.05%     
==========================================
  Files          41       41              
  Lines        6397     6347      -50     
==========================================
- Hits         5905     5856      -49     
+ Misses        492      491       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[anakryiko]

It feels like too much effort just to test this condition, especially taking into account all the "Ugh" remarks ;)

If you absolutely want to have a test for this specific conditions, can you use some root-only file as a source. E.g., /proc/kcore (assuming it's Linux and kcore is present, of course), or /etc/sudoers (seems to not allow reading for non-root).

[d-e-s-o]

It feels like too much effort just to test this condition, especially taking into account all the "Ugh" remarks ;)

It does, but the effort has been invested by now. The test is entirely optional and won't run by default (though it will run in CI, but that's a pretty stable environment overall and we can always disable it if it turns out to be flaky).

If you absolutely want to have a test for this specific conditions, can you use some root-only file as a source. E.g., /proc/kcore (assuming it's Linux and kcore is present, of course), or /etc/sudoers (seems to not allow reading for non-root).

The problem is more that all tests are running as root so we will always have access to these files. That's why the fork dance is necessary, so that we can run as different user. The main reason for checking this condition was that I wanted to test the behavior that Salvatore described. There is admittedly no coverage increase, but I'd say it's still valuable to have and risk and maintenance burden seem low to me. Let me know what you think.

[anakryiko]

Sure, why not. We can always delete it if it causes problems in the future, of course.

[d-e-s-o]

A bunch of the nastiness evaporates once we move this test into a separate binary.