Fix for #3859

[ehiggs]

Stop ignoring .gitnore files. Add a failing test which now passes..

[rust-highfive]

r? @brson

(rust_highfive has picked a reviewer for you, use r? to override)

[alexcrichton]

@bors: r+

Thanks!

[bors]

📌 Commit c21fff6 has been approved by alexcrichton

[bors]

⌛ Testing commit c21fff6 with merge d4d039f...

[bors]

💔 Test failed - status-travis

[ehiggs]

Welp, it fails on my test but I'm puzzled how it works on almost every platform but one. Possibly a timing issue with flushing to disk? Or is that just a hunch from working with too many broken parallel file systems?

[alexcrichton]

Hm that does seem odd, but it looks like Travis on the PR is also failing for the same reason? It looks like it may be related to a race though as the builder failing is changing?

Perhaps some more debugging information could be added to help diagnose?

[ehiggs]

Sorry, I think I wasn't clear. It is failing on travis and the test that fails is the new code I added. I can't reproduce it locally on linux or OS X using eatmydata (which is a nice program which makes fsync into no-ops). But I do see failures with all the tests using templates if I just run them as ./target/debug/new-.... so maybe that's a thread to pull on.

[alexcrichton]

Interesting!

In the face of #3878 I think we may want to close this though, right?

[bors]

☔ The latest upstream changes (presumably #3878) made this pull request unmergeable. Please resolve the merge conflicts.

[ehiggs]

With templates pulled out until we have an RFC that's passed, this PR is no longer valid.