Allow builds to fail

[0xADD1E]

In some cases, a build would fail (due to a particular malformed scheme) but we still want to be able to build the correctly formed schemes. To this end, this separates building a file into a fn -> Result. If one of the schemes fails, we log it and move on.

Additionally (to make it more clear if a scheme fails to build, some of the visual noise (messages for each created file) have been taken down to debug level log messages.

[ilpianista]

Hi, what base16-builder-php does when a scheme fails to build? I'd like to avoid to add a behaviour which isn't included in the builder specs.

May you please try to get this behaviour accepted in the specs, first?

[0xADD1E]

So, to address the two points...

The specification

The specs are (deliberately?) fairly vague with regard to the incredibly low-level details of builders. I don’t think error handling would disagree with the letter of the specification, and almost certainly not the spirit.

Comparison to builder-php

Builder-PHP has some different behaviour than this builder in the first place. The main one I noticed was the fact that builder-php shells out to git (which tracks branches nicer than libgit2), instead of doing the libgit2 checkout. This was generally the first reason I added the error handling, was that on subsequent updates, the branch to fetch was hard coded as origin/master, which in the case of the horizon scheme, doesn’t exist. For the rust builder this aborts, and is somewhat inconvenient.

I also haven’t gotten the opportunity to investigate the build failures themself, but it could be related to this. I’ll do more triage later today and write up a proper issue for the origin/master hard coding (particularly if I can figure out the libgit2 way of finding the upstream branch), and if it’s not related, the build failures

[peterkos]

This is cool! +1 that handling a not-working scheme we try to reference shouldn't crash the program.

I tried running against this branch and got the two following errors:

# target/debug/base16-builder update
thread 'main' panicked at src/main.rs:309:27:
Failed to clone: remote authentication required but no callback set; class=Http (34); code=Auth (-16)
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

# target/debug/base16-builder
thread 'main' panicked at src/main.rs:321:27:
Failed to clone: remote authentication required but no callback set; class=Net (12)
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Which seem to be lingering .unwraps() elsewhere, but then I got this one:

target/debug/base16-builder
thread 'main' panicked at src/main.rs:183:51:
called `Result::unwrap()` on an `Err` value: Os { code: 2, kind: NotFound, message: "No such file or directory" }
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace