From: Daniel Luz Date: 2008-07-25T11:53:43+09:00 Subject: [ruby-core:17953] Re: We'll release 1.8.6/1.8.7 this Friday ------=_Part_105249_19989375.1216954777676 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Content-Disposition: inline On Thu, Jul 24, 2008 at 9:19 PM, Nobuyoshi Nakada wrote: > SUS defines nothing about the conversion detail, nor encoding > names. Some implementation assign different encoding names for > UTF's with/without BOM, but it's definitely implementation > specific. I think the spec depends too much on particular > platforms. > You're right. Strictly speaking, the iconv library seems completely impossible to test, because there's not a single encoding which is guaranteed to exist across all implementations, much less behave consistently. So, when writing those specs, I tried to find a minimum of encodings which would be available on any implementation from the last couple years (the most common Unicode-related ones), and then tried to use the least common denominator I could find among some platforms (basically, whatever libraries were being used by the common Ruby distributions on Windows, OS X and Ubuntu). This, of course, could still fail on some implementations, so feel free to remove those problematic tests. ------=_Part_105249_19989375.1216954777676 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Content-Disposition: inline
On Thu, Jul 24, 2008 at 9:19 PM, Nobuyoshi Nakada <nobu@ruby-lang.org> wrote:
SUS defines nothing about the conversion detail, nor encoding
names.  Some implementation assign different encoding names for
UTF's with/without BOM, but it's definitely implementation
specific.  I think the spec depends too much on particular
platforms.

You're right. Strictly speaking, the iconv library seems completely impossible to test, because there's not a single encoding which is guaranteed to exist across all implementations, much less behave consistently.

So, when writing those specs, I tried to find a minimum of encodings which would be available on any implementation from the last couple years (the most common Unicode-related ones), and then tried to use the least common denominator I could find among some platforms (basically, whatever libraries were being used by the common Ruby distributions on Windows, OS X and Ubuntu). This, of course, could still fail on some implementations, so feel free to remove those problematic tests.
------=_Part_105249_19989375.1216954777676--