[Clfs-dev] Latest git build Pure 64 systemd bison and flex

Martin Ward macros_the_black at ntlworld.com
Fri Apr 4 00:24:48 PDT 2014


On 04/04/14 01:16, William Harrington wrote:
>
> On Apr 3, 2014, at 5:30 PM, Martin Ward wrote:
>
>> Hi
>>
>> I notice that bison and flex have been moved before binutils in both 
>> the cross-compile and final system stages, unless i'm missing 
>> something completely obvious, release tarballs don't need these 
>> programs and never have done. I can't find anything about ar linking 
>> to the flex library anywhere as William stated in his post to the 
>> list earlier this month.
>>
>
> Flex 2.5.38 and 2.5.39 install shared libraries. Without flex in front 
> of binutils in final system, ar links to the shared library in 
> /tools/lib or /tools/lib64 and when building gcc in final system, it 
> will error because the shared library is not in /lib or /lib64 but 
> binutils found the flex library in /tools and linked to it.
this is the point, i can't find any evidence to support this, on my 
system it doesn't
>
> Of course fsf binutils doesn't need bison or flex in front of it, but 
> Chris and I have been doing this recently because of it.
>
> We could disable shared library in temp system flex, but then binutils 
> ar will link against temp system libfl static library.
> It is probably a good idea for ar to link with a flex library in final 
> system. Static or shared.
>
>>
>> Obviously they should remain in the final stage but it seems better 
>> to build flex before bison as this removes the need to pass
>>
>> ac_cv_prog_lex_is_flex=yesto configure but it does have ramifications 
>> to the flex test suite, but considering two flex tests involving 
>> don't work anyway, seems a good idea to me, opinions?
>>
>
> This is probably worth looking into. I'm about to run a build with 
> tests and will look further into it.
cool
>
> Sincerely,
>
> WIlliam Harrington
>
>
martin



More information about the Clfs-dev mailing list