Re: [Gtk-osx-users] Symbol not found: _iconv on LION



On Jul 21, 2011, at 1:45 PM, John Ralls wrote:

> 
> On Jul 21, 2011, at 1:11 PM, John Ralls wrote:
> 
>> 
>> On Jul 21, 2011, at 9:22 AM, John Ralls wrote:
>> 
>>> 
>>> On Jul 21, 2011, at 2:26 AM, Richard Procter wrote:
>>> 
>>>> 
>>>> On 21/07/2011, at 3:53 PM, Olivier Sessink wrote:
>>>> 
>>>>> Hi all,
>>>>> 
>>>>> I've received various reports from users that have upgraded to Lion that 
>>>>> Bluefish won't start anymore:
>>>>> 
>>>>> Dyld Error Message:
>>>>> Symbol not found: _iconv
>>>>> Referenced from: /usr/lib/libcups.2.dylib
>>>>> Expected in: 
>>>>> /Applications/Bluefish.app/Contents/Resources/lib/libiconv.2.dylib
>>>>> in /usr/lib/libcups.2.dylib
>>>>> 
>>>>> I don't really understand this message, /usr/lib/libcups.2.dylib 
>>>>> references a function in 
>>>>> /Applications/Bluefish.app/Contents/Resources/lib/libiconv.2.dylib ??
>>>> 
>>>> Not directly, but if the dynamic linker is configured to prefer your bundled libiconv then that is what it will be linked against at runtime. See dyld(1) and DYLD_LIBRARY_PATH. 
>>> 
>>> Unfortunately, not so easy. Once the libiconv in the bundle is loaded, dyld won't load any other ones, so you still get the error from libcups.
>>> 
>>> Having dl'd the sources for Apple's libiconv, it appears to have diverged substantially from gnu's. I originally included libiconv into the gtk-osx build because of problems with x86_64; that turns out to have been the wrong solution.
>>> 
>>> I think the solution is to remove libiconv and rebuild everything; the automake macro which (incorrectly) decides that it's gnu-compatible will have to be fixed.
>> 
>> Spoke too soon: Attached is a patch to libiconv that corrects the missing symbols. It's not workable upstream as-is, though, so don't go filing a bug against libiconv with that patch. I still think the problem is better addressed by forcing everything to work with iconv_foo instead of libiconv_foo if that's achievable.
>> 
>> Unfortunately for Gnucash and anything else that uses it, there's another crash in libgdk-quartz. I need to debug it a bit more and then file a bug against Gtk; I'll report the bug number when I have it.
> 
> Sorry, forgot the attachment.

And the bug is https://bugzilla.gnome.org/show_bug.cgi?id=655078. I even figured it out, and I'll patch gtk-2.24-quartz and quartz-integration soon.

Regards,
John Ralls


Reg



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]