[boost] Re: boost::crc and linking problems

From: Lars Gullik Bjønnes (larsbj_at_[hidden])
Date: 2002-10-31 06:04:36

Daryle Walker <dwalker07_at_[hidden]> writes:

| On Thursday, October 24, 2002, at 10:04 AM, Lars Gullik Bjønnes wrote:
>
>>
>> Did you see my message to the list?
>> Do you have any idea of a less hackish solution?
>>
>
| I saw the message.  It sounds like a problem with how to use the
| tools, or how the tools work with each other.  The code itself isn't
| the problem, so I don't know how to fix it.

Are you _sure_ this is not a problem with the code?

This problem does only exist for Boost.crc not for any other Boost lib
(that we use.)

Regardless if this is a tools issue or not, of course it might be a
compiler/linker bug, it would be _really_ nice to resolve it in the
lib. So that "magic" hacks in the application code could be avoided.

>> This is a comment in our code currently:
>>
>>     // OK, this is ugly, but it is the only workaround I found to
>> compile
>>     // with gcc (any version) on a system which uses a non-GNU
>> toolchain.
>>     // The problem is that gcc uses a weak symbol for a particular
>>     // instantiation and that the system linker usually does not
>>     // understand those weak symbols (seen on HP-UX, tru64, AIX and
>>     // others). Thus we force an explicit instanciation of this
>> particular
>>     // template (JMarc)
>>     template class boost::detail::crc_table_t<32, 0x04C11DB7, true>;
>>
>> Do any of you have ideas for a better solution?
>

-- 
	Lgb