Mingshen Sun

Hi, @pluth, thanks for your report. @lizhuohua and I will look into this issue and get back to you later. Thanks.

> The coverage significantly dropped after the tests were refactored (and it’s clearly wrong right now as the parser for sh says 0%). This is wired and needs more investigation...

True. Please ignore my previous simple implementation. We need a decent `getty`.

Still zero coverage for `parser.rs`: https://codecov.io/gh/mesalock-linux/mesabox/tree/373875043044f1f248f48477d3cf19f92fab8273/libmesabox/src/posix/sh

I found that the `libmesabox` test (unit tests) reported coverage statistics of `parser.rs` (though zero coverage because of no unit tests for parser I guess). However, if we only look...

Another case is that the `parser_num` function is explicitly called by tests, but the coverage results show it was not hit. https://codecov.io/gh/mesalock-linux/mesabox/src/master/libmesabox/src/posix/head.rs#L339

With further investigation, I found that functions in `validator_os` in `clap` will not be counted. - https://codecov.io/gh/mesalock-linux/mesabox/src/c5674c4e02ddeb5a317e4a19a58c137df0524cc5/libmesabox/src/gnu/base32/common.rs#L142

0.4.16 has been released with the fix: https://github.com/alexcrichton/tar-rs/commit/b3ea06fcaf8b1bf38b4266da3e73bb43fbcbc2d5

I'll leave this issue open. When a consistent error handling method and guideline is ready, we can close this issue then.

> derives Fail (such as ChmodError) and then map errors from the standard library or outside libraries to that type Using `failure` is ok for me. For consistency, I found...