Get it working for JRuby and Rubinius #31
Comments
It's not guaranteeed to be in stdlib, and Syck fucks up the SeeingIsBelieving::Line serialization (turns it into an array) Missing Psych was causing rbx to fail, but now depending on it is causing JRuby to fail. Not sure what to do here. This is part of #31
Rbx passes all specs, and looks like feature issues are due to different presentation style of backtrace JRuby has one failing spec, but gets totally destroyed in the cukes b/c it's interface to Open3.capture3 is apparently different (looks like it ultimately delegates to IO.popen, but haven't tracked that code down yet This is part of #31
|
I think to get it passing on Rbx, I need to update Haiti to allow for more permissive matching in the backtrace. Might also need to deal with #32 |
No need to waste Travis's computers on these since I know I expect them to fail for now. And I'm not working on them as I'm putting #31 on hold for a bit. Also removing it from things to do for #47. JRuby has an IO bug at the moment, it's causing EventStream tests to lock up. jruby/jruby#2461 I could probably get around it by commenting them out, it's not for a case that I actually expect anyway. But 9k Fixes this anyway, so wait for that to be common. Rbx had a bug with how it was handling mass assignment. The "rest" variable on this line https://github.com/JoshCheek/seeing_is_believing/blob/2ddca5d42c693e3028d574dcb74bcc1f654b068c/lib/seeing_is_believing/binary/remove_annotations.rb#L79 is nil when thre is only one subarray. I had started to open a bug report, but realized m Rbx was a bit older. So was going to upgrade and see if its still there, maybe try and fix it. But I couldn't get it to install. Couldn't find llvm-config. I'm on an older OS (Lion) since everyone had issues going to Yosemite, and don't feel like going through the headache of upgrading just to get a newer llvm, just to get Rbx. Plus apparently Apple is now tracking a lot more data, and you have to run some Python script to turn it off or something, and IDK, this one is working sufficiently fine for me. Oh, but I wound up uninstalling Rbx while attempting to upgrade, so now I don't even have one that I can test against. I'll probably come back to these when I have more time and it takes less effort.
|
sigh still a PITA :/ |
|
What's the problem on JRuby? I'd like to help. popen and spawn should behave like MRI now, with only a few behavioral differences. |
|
I attempted to bundle install, but you appear to have binding_of_caller as a dev dependency, and it did not appear to install on JRuby (C ext). |
|
Aye, that's a dep of |
|
I think these were the issues I had the last time I put a relatively serious amount of effort into getting it working: https://github.com/jruby/jruby/issues?utf8=%E2%9C%93&q=is%3Aissue%20author%3AJoshCheek but don't trouble yourself with it, I'll take a look Note that I just tried, and it looks like I may have to remove the vendored Bundler (cuts the runtime significantly since I'm running a lot of subrocesses). Anyway, I'll check it out further tomorrow. |
Plus some other changes while exploring issues blocking #31 Current issue looks like the file isn't there when it goes to delete it. There's about a hundred, many are different, but all could feasible be caused by this
They're close! Actually, Rubinius should work, it just doesn't pass all the tests. JRuby is broken, though, interface to
IO.popenseems different.The text was updated successfully, but these errors were encountered: