Revise error types and codes

[dhardy]

This makes a few changes:

• revises error codes to have a common prefix
• allows custom codes and messages per target
• makes the error module public

It feels quite clunky having to add an empty fn error_msg_inner to all modules, but I don't see a better approach.

@newpavlov review? I think this concludes all my to-do list for the first release.

BTW the dummy module was broken; we don't appear to have a test for it (I hacked lib.rs locally).

[newpavlov]

Overall looks good to me! Though I think we do not need error module and we can re-export Error type to the crate root.

Also looks there is a typo in Error::code docs:

This may equal one of the codes defined in this library or may be a system error code.

Fix to "this may be equal to one of <..>"? Also shouldn't "it" or "the return code" be used here instead of "this"?

[newpavlov]

Maybe we should make UNKNOWN and UNAVAILABLE associated constants of Error type?

[dhardy]

We also don't need the std::error module, but I don't think it gets in the way? At least in this case the other error types can have better names.

[newpavlov]

I think a better analogy will be std::io::error, which as you know does not exist. Think in terms of potential inclusion into std.

[dhardy]

And yet std::io::prelude exists, while there are several error-related things in std::io. But okay, I guess we can do away with the error module.

[newpavlov]

BTW maybe it's worth to add UNKNOWN and UNAVAILABLE error code values to relevant constant docstrings (in both hex and dec forms), so it will be possible to search for them via search engines?

[newpavlov]

Sorry, I meant "Error(0x{:08X})" in the comment earlier. Also what do you think about associated constants? And maybe we should use 24-bit prefix for easier search? For example Ox57f4c5OO results in zero google hits.

[dhardy]

Okay. I added the mentioned logging. What do you think about this? Many of the messages are not hugely useful, but this feature is optional and it's still not much overhead.

[newpavlov]

No objections, though I am also not sure about usefulness, as you either get a panic, or process error explicitly.

[dhardy]

Agreed. Still, I don't wish to spend the time on trivial tweaks now so this will do (once I fix tests).

[dhardy]

FYI, libc version bumped to solve this error. I lack motivation to find the minimal version required.

[newpavlov]

I can prepare a PR with some minor tweaks after you'll merge this one.

[dhardy]

Please do.