search

dkubb / adamantium

Create immutable objects
MIT License
373 stars 13 forks source link

MRI 2.1.0-p0 segfault #30

Closed ghost closed 10 years ago

ghost commented 10 years ago

Occurs in Adamantium::ClassMethods#new Core dump: https://gist.github.com/lgierth/8411173

Getting this with both 0.1.0 and current master (7076723698b1).

I'm figuring out whether this has already been reported to MRI, there are a couple of recent segfault-related tickets, but they seem unrelated to this one.

ghost commented 10 years ago

On Travis as well: https://travis-ci.org/lgierth/promise.rb/jobs/16904962

ghost commented 10 years ago

Ah, it's fixed in 2.2.0dev (2014-01-12 trunk 44563), so I should probably just wait for 2.1.1 to be released.

mbj commented 10 years ago

@lgierth For reference. 2.1.0 crashes also with anima.

dkubb commented 10 years ago

@lgierth I'm curious if you tried any work-arounds, like maybe restructuring the code to not trigger the exception? I would accept patches to change the code as long as it didn't change the functionality.

I may do some digging too, since this is causing failures on a few other projects I'm testing on 2.1.0. I'd rather have something working now than wait for 2.2.0.

mbj commented 10 years ago

@dkubb For you interest: I found numerous of segfaults with MRI-2.1.0, in various places of my commercial code base. I'd not use 2.1.0 in production. Seems this release wasnt that well tested. Hoping for next.

dkubb commented 10 years ago

@mbj I'm kinda curious if the MRI release is ever run against the rubygems corpus. It seems like it would be easy to get a list of gems that pass with a previous release (from travis or circle-ci), and then run the test suite against them. I'm guessing the travis or circle-ci would even help with such a test, especially prior to a stable release.

mbj commented 10 years ago

@dkubb From my discussions with implementators of jruby/rbx I remember that the MRI guys dont even use rubyspec during their build.

I'd love if MRI would listen more and cleanup ugly issues in ruby instead of adding more and more syntax.

ghost commented 10 years ago

@dkubb Didn't try variations of the code, since I also have a couple of other problems with 2.1.0, and thus just decided to consider it a half-broken release.