tech-pkg archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Problems building devel/subversion-base on MacOS X 10.8

Am 29.07.2013 um 11:09 schrieb OBATA Akio <>:

> On Mon, 29 Jul 2013 17:39:04 +0900, Jens Rehsack <> 
> wrote:
>> Am 26.07.2013 um 15:27 schrieb Joerg Sonnenberger 
>> <>:
>>> On Fri, Jul 26, 2013 at 09:19:17PM +0900, OBATA Akio wrote:
>>>> ${PREFIX}/lib/charset.alias is always used in groff, independent of whether
>>>> builtin libiconv or not, because groff itself is not using libiconv.
>>>> So I feel that hard dependency on converters/libiconv is right.
>>> I wonder if we should flag this as bug upstream. I don't see why groff
>>> should do canonisation of character set names in first place.
>> Well, but for the moment we need some kind of fix.
>> If Akio would use, usual result would be applied.
>> That might allow overriding of USE_BUILTIN.iconv for NetBSD's without 
>> charset.alias.
>> Further: it looks that it searches for libiconv - so they might know about 
>> the
>> conflicting charset.alias which is tried to prevent by patch-ab …
>> Simply patching the own charset.alias out and adding a hard requirement to
>> converters/libiconv is no solution because of the side effects.
>> Other proposals?
> Probably, origin of this issue is same as PR pkg/48036, it should be fixed.

Yes, hidden dependency. We can fix that only by intercepting the generated *.la
files. I have no clue how we can do that.

> Dropping dependency on converters/libiconv for groff is not relevant to
> this issue.  Even if the dependency is dropped, the issue will still happen
> with existence of iconv from pkgsrc.

Nope - it happens because of existence of iconv from pkgsrc and was
linked/referenced by the - what is even more tricky.

But I would also argue, that the nit (from weight of error) in textproc/groff
should be fixed, anyway. But I would prefer a kind of fix making you and me
happy ;)

Jens Rehsack
pkgsrc, Perl5

Home | Main Index | Thread Index | Old Index