Add .gitignore for new packages #9541
Conversation
|
Hmm... looks like a timeout on Windows and failure to download libgit2 on OSX. |
|
I'd also put |
|
lgtm. it's optional whether you do them in separate files or not. i disagree that |
|
I'd maybe hold off on adding Coverage/Coveralls.io until JuliaCI/travis-build#1 is resolved |
|
Okay, edit to remove |
|
Oh hold on, I just realized all those lines are commented out. Yeah it does seem to be one or the other... maybe needs additional comments to explain what all that Coverage stuff is. |
|
Additional comments would be helpful -- not sure where to go with them before JuliaCI/travis-build#1 is resolved, as you said -- guess I should update this then. Also, thoughts on whether the commented out code ought to match the default script or be a useful alternative (e.g. opposite of the default script in terms of doing code coverage)? |
|
The idea of the default script on the travis side is to reflect package-testing best practices. Having the commented-out section there in the generated |
| @@ -149,6 +150,17 @@ function travis(pkg::AbstractString; force::Bool=false) | |||
| #script: | |||
There was a problem hiding this comment.
To clear things up, and since people switching over to the default script are mistakenly thinking that coverage=true is set, how about adding this line here for now?: # (the following snippet differs from the default test script in that coverage/Coveralls are not enabled by default)
There was a problem hiding this comment.
I feel like more of these descriptions should be going in the package development documentation (based on feedback I've seen various places, people seem to be missing that it exists, or it needs to be expanded to cover more use cases) rather than comments in a generated file.
Also I guess I was hoping JuliaCI/travis-build#1 would be resolved a little sooner, in that at least coverage=true could be made the default on Travis before we figure out the rest of it.
There was a problem hiding this comment.
That sounds better -- I reverted the script comment change so it again ends with coverage=true) in anticipation of the change. Thanks for the suggestions.
The clutter seems worth it to make code-coverage and track-allocation more painless (/ hopefully common).
garborg
changed the title
|
win32 appveyor failure is unrelated (@vtjnash should really work through pull requests more often to avoid breaking master), travis osx failure is Homebrew/legacy-homebrew#32819 |
Add .gitignore for new packages
|
does coverage / track-allocation work on 0.3? if so we should label for backport pending |
|
This could be backported |
|
backported in 30a892b |
Adding the .gitignore: the clutter seems worth it to make code-coverage and track-allocation more painless (/ hopefully common)
Completing Travis script: still somewhat confusing, as the commented script is not the default script, but so long as it's not, might as well make it a complete version of the most commonly used alternative to the default