language version markers

[graydon]

We need a way (using the hash-comment "pre-parse" form similar to shebang comments) to mark a crate's (symbolic) language version, so we don't try parsing rust when we don't have a parser that can handle it. We'll mandate the presence of this sort of marker on a packaged-for-distribution rust file.

[pnkfelix]

This need not block 0.6.

[graydon]

nominating for backwards compatible

[graydon]

accepted for backwards-compatible milestone

[emberian]

Visiting for triage; nothing to add

[brson]

Nominating. This ought to be done with #3795.

[pnkfelix]

If you want fully-general language versioning, then the approach of #3795 does not seem sufficient.

(In particular, you still need to be able to parse in order to deal with the #[cfg(..)] attributes, so how would you deal with parser changes...)

[pnkfelix]

(we're going to leave I-nominated so we can wait until next week and see if @brson can explain his thoughts further.)

[pnkfelix]

((it seems like some of the discussion on #3795 elaborates further on what @brson might have been thinking. E.g. @cmr said there that one should only attempt to work at the level of per-crate granularity, which would probably side-step the parser issue, though I think that per-crate may be too coarse -- per mod-in-a-file is more what I was thinking...))

[brson]

P-low, not 1.0

[brson]

If something ever happens here it will be done as-needed. It's not really on the horizon right now. Closing.