Comments on: Ruby GEMS still doesn’t work on 10.4?

By: AdaRuby: Web 2.0 Development with Rails » Fixing RubyGems in Ubuntu Gutsy Installation

Mon, 24 Mar 2008 13:26:23 +0000

[...] VMUNIX Blues » Blog Archive » Ruby GEMS still doesn’t work on 10.4? [...]

By: ruby, gcc, and powerPC macs - Todd Pinkerton

ruby, gcc, and powerPC macs - Todd Pinkerton — Sat, 07 Apr 2007 02:54:33 +0000

[...] It turns out, ruby compiled wth gcc 4.0 on powerPC macs — which my laptop is one — is known to exhibit these strange behaviors. David Heinemeier Hansson himself mentioned this in this blog post. [...]

By: mark

mark — Fri, 14 Apr 2006 06:30:58 +0000

Well my first attempt at using “gcc_select 3.3″ was rather unpleasant. You see, I have a lot more than Ruby compiled in /usr/local, and it was all built with GCC 4.X. Right off the bat after I rebuilt Ruby (and re-gem’ed) with GCC 3.3 any attempt to run a Rails app got me a dynamic loader error message relating to libreadline. So I ended up rebuilding readline, mysql, and sqlite with gcc 3.3. Then I realized two of my C apps that were built with gcc 4 broke; now they couldn’t link to libmysql. Yay.

Now, you see, sometimes I have to do some heavy compute stuff on my G5. And GCC 4 gives me autovectorization of code via the -fast switch. It makes a HUGE difference on some of my code. Like, the code runs *several times faster*. So recompiling those programs with gcc 3.3 simply isn’t an option.

So I’m sort of stuck. Why can’t ruby just work with gcc 4? Pretty please?

By: mark

mark — Mon, 10 Apr 2006 04:27:54 +0000

I actually don’t see the problem on my powerbook g4, but do on the dual g5. I’ll try the gcc 3.3 switch, but has anyone actually figured out what the gcc4 problem is? Everything else I run benefits greatly from gcc4, with the autovectorization stuff in particular doing wonders on some of the scientific apps. Kind of sad to have to go back to gcc3.

By: David Heinemeier Hansson

David Heinemeier Hansson — Mon, 10 Apr 2006 00:27:03 +0000

G4 macs have trouble with GCC 4.0. Use gcc_select 3.3 to switch to 3.3, recompile Ruby, and rejoice.

By: mark

mark — Sun, 09 Apr 2006 08:01:00 +0000

Faisal: Yeah, the second time almost always goes through fine for me. It’s often the RDoc part where it blows up, so I have to come back and specifically keep trying the RDoc install until they all succeed.

Personally for me this rarely happens for me on my PowerBook, but happens all the time on my Dual G5. I’m using the very latest gcc 4 on both.

I have yet to see it happen on Linux, but on Linux I am using gcc 3.4. I thought I remembered when the problem was getting some attention before the Ruby 1.8.4 release that people were seeing the Fixnum thing on both gcc 3.4 and 4.x, though..

By: Faisal N. Jawdat

Faisal N. Jawdat — Sun, 09 Apr 2006 01:16:12 +0000

Two things that seemed to help:

1. Keep trying. For some bizarre reason this (and other) problems seem to go away when trying again (and again and again and again).

2. Set the CC environment variable to gcc-3.3 — 4.x seems to have some problems.

By: Dave Duchene

Dave Duchene — Sun, 09 Apr 2006 00:56:38 +0000

gems is working fine for me, but I do see a heisenbug very similar to that when developing with Rails and Webrick. Every once in a while, I get an unhandled exception screen, unknown method foo for -largenumber:Fixnum. Hitting reload in the browser fixes it. I’ve seen it happen on Debian Linux on Intel and OS X on a G4. I wonder if our problems could be related?

By: jwa

jwa — Sat, 08 Apr 2006 22:06:48 +0000

I had no problem upgrading on my Mac, so I don’t think it’s GEMs, but something about your machine.