Erik Huelsmann wrote:
>>On Thu, 2004-05-13 at 14:13, Erik Huelsmann wrote:
>>
>>
>>>In order to prevent charset conversion by 'smart' gettext
>>>
>>>
>>implementations
>>
>>
>>>our build system has to strip out the the 'charset=UTF-8' string in the
>>>administrative section of po files.
>>>
>>>
>>So, I recall Nico repeatedly pointing out that all the gettext
>>implementations which perform charset translation also support the
>>function call to turn it off. I don't recall seeing an answer to this
>>claim.
>>
>>
>
>I don't think there was one. He stated that he thought they all did.
>
>
>
>>Given that stripping out the charset directive yields ugly warnings and
>>is presenting a portability problem, why are we doing it when there's a
>>better option?
>>
>>
>
>It's the only way that I currently know of (given the absence of reactions
>to state that he is correct) to be sure to eliminate the recoding. OTOH, if
>we are willing to give it a try to use the bind_textdomain_codeset() when
>available and assume a 'dumb' gettext if not until proven wrong that's fine
>by me.
>
>It was Branko who insisted on not having it in our code at all; I think he
>expected problems with the custom built GNU gettext. Maybe now that he has
>built it he can give us an answer to the question whether we still can't use
>it or not... (Branko, any idea?)
>
>
I was told on this list that the Solaris gettext didn't have
bind_textdomain_codeset. The warnings Greg mentsions are, as far as I
know, from GNU msgfmt. But this is all hearsay.
If indeed bind_textdomain_codeset is a GNUism, and if we can avoid the
need to use it by stripping the charset bit from the .po files /and/
expect all gettext implementations to behave in the same way afterwards
(i.e., not try to translate anything), then let's do that.
-- Brane
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri May 14 00:11:59 2004