From: Andrew Walrond <andrew@...>
Date: 2009-02-12T00:34:39+09:00
Subject: [ruby-core:21990] Re: ruby 1.9.1 parallel make race

Andrew Walrond wrote:
>
> Here's a patch against 1.9.1-p0 (also attached):
>
> --- ruby-1.9.1-p0/lib/mkmf.rb   2009-01-28 09:23:46.000000000 +0000
> +++ ruby-1.9.1-p0-apw/lib/mkmf.rb       2009-02-10 23:21:51.510922231 
> +0000
> @@ -1703,7 +1703,7 @@
>       for f in files
>        dest = "#{dir}/#{File.basename(f)}"
>        mfile.print("install-rb#{sfx}: #{dest}\n")
> -        mfile.print("#{dest}: #{f}\n")
> +        mfile.print("#{dest}: #{f} #{dir}\n")
>         mfile.print("\t$(#{$extout ? 'COPY' : 'INSTALL_DATA'}) ")
>        mfile.print("#{fseprepl[f]} $(@D#{sep})\n")
>        if defined?($installed_list) and !$extout
>
> (applies successfully against TRUNK also)
>
UPDATE

While this patch is valid, I have discovered other problems with 
parallel building which can result in build failure and, more 
worryingly,  a successful build but a broken ruby.

Symptoms I have seen:

'generating callback.h' appearing 3 or more times in a row, followed by 
lots of warnings

Successful build but borked libraries

/bin/ruby: symbol lookup error: /lib/ruby/1.9.1/x86_64-linux/dl.so: 
undefined symbol: rb_dl_init_callbacks_0

I'll try to come up with some more patches.

Andrew Walrond