Output columns 1-based. Fixes #10848

[jhasse]

No description provided.

[alexcrichton]

I agree that we shouldn't really be diverging from what standard tools like gcc are doing, but I'd like to at least see a test or two when dealing with this. I would also like to try to fix this at the source because anyone inspecting a Span would have to remember that the line numbers start from 1, but the column numbers start from 0. It'd be much nicer if everything either started from 0 or started from 1, and for ease of display it sounds like everything should start from 1.

[jhasse]

As @huonw said in #11183: "storing them 0-based makes indexing/slicing the strings by adding to the start-of-column offset more natural.", so maybe only outputting them 1-based is a good idea?

Tests would definitely be nice, but I'm not familiar enough with Rust to write them. Should I close this PR?

[huonw]

A perfectly ok test would be:

# //~ ERROR 1:1: 1:2: error: expected item

placed in src/test/compile-fail/column-offset-1-based.rs (you can then check this test passes with make check-stage2-cfail TESTNAME=column-offset).

[brson]

I agree that storing them 0 based and only incrementing for display is ok. @jhasse the test @huonw describes will work; just add a followup commit to this PR with the test case.

[jhasse]

Done, thanks!

[jhasse]

Damn I did git commit --amend instead of git commit -a --amend after fixing a typo. The test will fail, sorry! New patch should work.