curl code coverage daniel.haxx.se/blog/2023/03/28…
Let me explain why we don't have code coverage numbers for #curl.
⇧
curl code coverage daniel.haxx.se/blog/2023/03/28…
Let me explain why we don't have code coverage numbers for #curl.
Gen X-Wing
in reply to daniel:// stenberg:// • • •daniel:// stenberg://
in reply to Gen X-Wing • • •Gen X-Wing
in reply to daniel:// stenberg:// • • •*phew*
With the massive amount of architectures etc, I almost wonder if running the tests in some form of “virtual C interpreter” that doesn’t represent an actual computer could be a thing. Wouldn’t be surprised if there’s some static analysis startup at some point in time that had that idea.
Either way, full agree on the practicality of it all. My own personal way of viewing it tends to be “start out with the theoretically best and most correct, then move towards practical”:)
Timo Tijhof
in reply to daniel:// stenberg:// • • •Other fallacies on coverage:
Even if you combine all coverage reports and get ~100% test coverage, it could still be broken on Windows/macOS. One "has" to combine in order to reach code that is specific to certain builds/platforms. But, that implies covering it once is enough which misses the point of having different builds.
Incentive to create tests that have net-negative value, yet increase coverage. Tests that observe internals, mock too much, yet make no meaningful assertion.