From: Roger Pack <rogerpack2005@...>
Date: 2008-11-23T08:18:52+09:00
Subject: [ruby-core:20034] Re: [BUG?] rdoc/ri on solaris 8 (i386 vs sparc)

I assume they both use the same version of rdoc?  Do they both
consistently output the same [yet different from each other docs]? How
does it compare to other systems?
-=R

On Wed, Nov 19, 2008 at 12:49 PM, Ben Walton <bwalton@artsci.utoronto.ca> wrote:
> Hi All,
>
> I'm attempting to package 1.8.7p72 for solaris as part of the
> opencsw.org/blastwave.org projects.  This sees the build process done
> on both i386 and sparc versions of solaris 8.  The build/install goes
> fine, but I'm hitting a strange issue when building the ri system
> docs.  The list of files generated is _different_ on the two
> platforms, which means I can't release a single ARCH=all rubydoc
> package.  The differences are slight (eg: only about 4 files [see
> below]), but enough to cause problems.
>
> I've been poking at this for a while now and can't seem to discover
> where the issue is creeping in.  The only differences in the 2 source
> trees seem sane (different macros and compiler options for the
> different platforms, but nothing that looks amiss).
>
> At this point, I think there must be some low-level issue between the
> string libraries and the tokenization that happens during the rdoc ->
> ri parsing...?
>
> The only other oddity that I've detected is that when running `make
> check`, there is a fatal error related to iconv (see below).  We're
> linking against libiconv 1.9.2 currently.  This error is consistent on
> both platforms.  From what I can see though, iconv isn't loaded at all
> during the rdoc run (as evidenced by spitting out $" at the end of the
> run...).
>
> ---snip---
>  1) Error:
> test_translit_option(TestIconv::Option):
> Iconv::InvalidEncoding: invalid encoding ("SHIFT_JIS//TRANSLIT",
> "EUC-JP//translit//ignore")
>    ./test/iconv/test_option.rb:25:in `initialize'
>    ./test/iconv/test_option.rb:25:in `new'
>    ./test/iconv/test_option.rb:25:in `test_translit_option'
>
> 2028 tests, 1346436 assertions, 0 failures, 1 errors
> ---snip---
>
> This is the output from both platforms after running `make rdoc` in
> the build directory:
> Files:   549
> Classes: 1630
> Modules: 835
> Methods: 8454
>
> So that looks fine, but if I list all files in the .ext/rdoc
> directory, sort the list and then compare it with the list from the
> other platform, I see:
>
> $ diff i386.txt sparc.txt
> 1774d1773
> < ./IRB/Context/_set_last_value-i.yaml
>
> Does anyone have some thoughts on this?  I'd appreciate any help.
>
> Thanks
> -Ben
> --
> Ben Walton
> Systems Programmer - CHASS
> University of Toronto
> C:416.407.5610 | W:416.978.4302
>
> GPG Key Id: 8E89F6D2; Key Server: pgp.mit.edu
> Contact me to arrange for a CAcert assurance meeting.
>



-- 
Thanks!
-=R