FI doesn't show when "Coverage has degraded!"

[evverx]

https://introspector.oss-fuzz.com/project-profile?project=systemd currently shows

and it isn't obvious that the coverage went down on 2025-05-13.

It would be great if it said when the coverage went down without having to go over the reports and compare https://storage.googleapis.com/oss-fuzz-coverage/systemd/reports-by-target/20250512/fuzz-dns-packet/linux/report.html
and https://storage.googleapis.com/oss-fuzz-coverage/systemd/reports-by-target/20250513/fuzz-dns-packet/linux/report.html

(It would also be great if it said that the same codepaths are still exercised and the totals went up. It would make it easier to figure out what kind of regression it is)

[DavidKorczynski]

@phi-go would you be able to assist here?

[phi-go]

Sure. @evverx does the corresponding plot not already cover your usecase? If you scroll down on the page the per-fuzzer plots are interactive and you can see that coverage went down from the 12. to 13. Though, I could add a link so that it scrolls to the plot? It probably is also helpful to expand the plot with covered and total number of lines based on the coverage report.

Regarding your second point evaluating whether the same codepaths are still exercised. This seems difficult to do correctly. Could you expand what you would like to have? I imagine you would like something comparing the lines/branches/etc between two days to get number of newly covered, no longer covered, and non-changed coverage elements. Though, this gets complicated when doing several days and correctly matching when there are code changes.

[evverx]

It's possible to scroll down (and that's what I did to figure out when the coverage went down) but I think it would be better if it was embedded in comments too (ideally with commit ranges or something like that because the goal is to figure out what the issue is and fix it if applicable/possible).

Regarding your second point evaluating whether the same codepaths are still exercised. This seems difficult to do correctly.

Agreed. In that particular case the ratio went from "1723/7263" to "1723/25162" so if it had been shown by FI somewhere it would have been a bit easier to make sense of the report.

[phi-go]

I see. Commit ranges are not available to my knowledge as they are not really exported by OSS-Fuzz except for the build log, this would be an issue for OSS-Fuzz first.

I should be able to add links to the coverage report before and after the degradation. Though, I would limit it to the last instance otherwise this can become noisy for highly fluctuating targets. The total numbers should be quick to add as well.

[phi-go]

@evverx I was working on a related issue so I added your suggestions as part of #2162