Da
Show enters and exits. Hide enters and exits.
| 01:13:29 | boyscout | Updated MSpec source to 60fa49cd. - 258aefb - Brian Ford |
| 01:13:29 | boyscout | Updated CI specs to RubySpec 2a1a307e. - 1ab8ad9 - Brian Ford |
| 01:18:54 | boyscout | CI: rubinius: 1ab8ad9 successful: 3241 files, 12123 examples, 36446 expectations, 0 failures, 0 errors |
| 03:30:15 | matthewd | brixen: On the offchance you're still around... are there tests for Rubinius::Config somewhere? If not, should there be? |
| 03:43:58 | rue | matthewd: Check under rbconfig? |
| 03:44:10 | rue | Or did you mean some rbx-specific parts? |
| 03:46:21 | matthewd | rue: Yeah, the latter |
| 06:10:28 | brixen | matthewd: there are no tests for Rubinius::Config |
| 06:10:33 | brixen | matthewd: it's generated |
| 06:10:38 | brixen | based on platform |
| 06:10:44 | brixen | not sure what the point of tests would be |
| 06:15:33 | matthewd | brixen: I mean the prims that implement it, not the values therein |
| 06:22:51 | matthewd | brixen: Trying to work out where to put specs for http://gist.github.com/331803 |
| 06:25:47 | brixen | for the parts in the vm, use vm tests like you have there |
| 06:26:05 | brixen | I don't see any ruby code in that gist |
| 06:26:18 | brixen | was there ruby code that had a bug? |
| 06:26:42 | matthewd | The first hunk is a change to a primitive... I couldn't see any VM tests covering prims, so assumed they were covered from specs |
| 06:27:16 | brixen | specs for primitives are based on the method wrapping the primitive |
| 06:27:33 | brixen | what method wraps that primitive? |
| 06:27:38 | matthewd | Yeah, that's what I figured. Which in this case, is Rubinius::Config.get |
| 06:28:02 | brixen | ok |
| 06:28:50 | brixen | should be spec/core/rubinius/config/get_spec.rb then I suppose |
| 06:29:15 | matthewd | I guess the trick will be the values, though |
| 06:29:18 | brixen | actually spec/core/config/get_spec.rb |
| 06:29:48 | brixen | since various things are under the Rubinius namespace, but doesn't really make sense to push them down a directory |
| 06:29:54 | matthewd | As you said, the values are platform specific... so it's hard to say what the spec can actually depend on |
| 06:30:06 | brixen | spec/core spec/library, etc are all Rubinius specific |
| 06:30:10 | matthewd | Short of having the config-generating stuff spit out some guaranteed values in a special test. namespace |
| 06:30:27 | brixen | there's no point making specs for the generated values |
| 06:30:50 | brixen | the expected values would be generated too |
| 06:31:02 | brixen | we don't control what the platform gives for a specific value |
| 06:31:26 | matthewd | Yeah, but to have a spec that says "parses a value of 123123123123123 as a bignum", we need some way of getting that number into the VM's config |
| 06:32:26 | brixen | huh? |
| 06:33:22 | brixen | again, I think this test belongs in the vm tests |
| 06:34:04 | matthewd | Okay, I think I'm confused |
| 06:35:58 | brixen | I am too |
| 06:36:08 | matthewd | There's no precedent of a call to vm_* in vm/test/... so it feels like this should be a spec (of the R::C.get method) |
| 06:36:19 | brixen | ok |
| 06:36:25 | brixen | why can't it be? |
| 06:36:29 | matthewd | But, the underlying data structure this primitive is reading is read-only |
| 06:36:49 | brixen | R::C.get("some number").should == some_bignum ? |
| 06:36:57 | matthewd | So, the spec can't put a dummy value in there (to parse, and then assert about) |
| 06:37:07 | brixen | then it needs to be a vm test |
| 06:37:17 | brixen | if you cannot provide it a value, how can you test it |
| 06:37:21 | matthewd | R::C.get("some.key.that.is.definitely.a.bignum").should be_kind_of(Bignum) |
| 06:37:22 | brixen | *what* are you testing even? |
| 06:37:39 | brixen | "how can the children learn if they cannot fit in the school???" |
| 06:40:44 | matthewd | I think I'm talking about two possibilities: one is that platform.conf is guaranteed to contain a line of "rbx.test.bignum = 999999999999", and the spec depends on that... the other, is yes, define it as a VM test |
| 06:40:46 | brixen | this seems easy to test with a vm test |
| 06:41:04 | matthewd | Okay, I'll do that. |
| 06:41:25 | brixen | I don't think it makes sense to have to make platform.conf contain a value |
| 06:41:43 | brixen | you can twiddle that data structure easily from a vm test |
| 06:42:19 | matthewd | Erm... assuming prims can be called from the middle of nowhere (don't expect anything special of their state)... I haven't looked that deep yet. |
| 06:44:40 | brixen | sure, this method is static |
| 06:44:49 | brixen | just a plain old static C++ method |
| 06:45:58 | matthewd | vm_get_config_item(STATE, String* var)... I was just wary of the STATE param |
| 06:51:31 | matthewd | brixen: Thanks. An offering to make up for annoying questions: http://gist.github.com/331810 :) |
| 07:06:13 | brixen | matthewd: cool |
| 07:06:39 | brixen | matthewd: lines 103-106 of that gist need work |
| 07:06:45 | brixen | generally, should_not is not allowed |
| 07:07:09 | brixen | if it calls #to_i on its argument, there should be a should_receive(:to_i) in that example |
| 07:08:04 | brixen | lines 155-157 should be in specs like: it "raises a TypeError when passed nil" |
| 07:08:07 | brixen | ... |