libtool-patches
[Top][All Lists]
Advanced

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

RE: AIX patch for libtool.m4


From: Boehne, Robert
Subject: RE: AIX patch for libtool.m4
Date: Fri, 30 May 2003 15:21:33 -0500

I'm not convinced because I have yet to see a piece of code
that doesn't work, post that and I'll believe!

Thanks,

Robert

-----Original Message-----
From: jedwards [mailto:address@hidden
Sent: Friday, May 30, 2003 3:00 PM
To: Boehne, Robert
Cc: Albert Chin; address@hidden
Subject: Re: AIX patch for libtool.m4


I tried to explain this to Robert before I don't think I've convienced 
him yet. 

Currently the default libtool behavior for AIX 5 is to do this with an 
archive:

cc -Wl,-bM:SRE -o .libs/libhello.so.2  .libs/hello.o .libs/foo.o  -lm 
-lc  -Wl,-bE:.libs/libhello.exp -Wl,-bnoentry
ar cru libhello.a libhello.so.2
ar cru libhello.a hello.o foo.o

which puts both these objects files in the archive twice - only the 
first entry into the archive is read so this is just a waste of space.  
Now I notice that there is an exception in libtool.m4 for aix4 which 
sorta fixes this problem by not doing the second ar command - you could 
just extend that to aix5.  But I think that its the wrong way to solve 
the problem.
The first ar statement, not the second - is the one which should be 
eleminated to produce a standard AIX library. 

Now for run time linking support AIX has the -brtl flag which causes it 
to look for a file libname.so before libname.a  I'm working on this 
change and I have almost everything going but its still failing a few of 
the tests -  I should be there soon. 

You are right about the template problem - thats what got me going on 
this whole thing in the first place. 

I'm probably not explaining things very well -
I hope Rob that you will give my change fair consideration when I get it 
done - including reviewing it with other IBM  authorities if you feel 
the need.



Boehne, Robert wrote:

>The only thing I was aware of is that archive members can't
>be dlopen'd by libltdl because the right flags aren't passed.
>
>BTW, would you post a link or reference to why -qmkshrobj is needed
>with templates when building SYSV-style libraries?  I wasn't aware
>of any problems with templates.
>
>Rob
>
>-----Original Message-----
>From: Albert Chin [mailto:address@hidden
>Sent: Friday, May 30, 2003 2:30 PM
>To: jedwards
>Cc: address@hidden
>Subject: Re: AIX patch for libtool.m4
>
>
>On Fri, May 30, 2003 at 01:20:44PM -0600, jedwards wrote:
>  
>
>>I'm working on a general update to libtool.m4 for aix5 and aix library 
>>support - do you want to wait a week or so to see what I manage to do? 
>>    
>>
>
>I think the only thing lacking is xlC support. Of course, there's also
>32/64-bit support but there's a workaround for that. What else are you
>working on?
>
>  
>
>>Albert Chin wrote:
>>
>>    
>>
>>>On Tue, May 27, 2003 at 02:40:04PM -0500, Boehne, Robert wrote:
>>>
>>>
>>>      
>>>
>>>>Ok, that was interesting.  Apparently Jim, you are correct, it is valid to
>>>>add -qmkshrobj to a -G command line.  However, I do not believe that
>>>>Libtool should do this by default.  If you want to add it yourself you 
>>>>can,
>>>>Libtool allows this with -XCompiler -XLinker and -XCCLinker.  If you still
>>>>think this should be done by default, let us know why.
>>>>  
>>>>
>>>>        
>>>>
>>>Wel, to get xlC C++ support for AIX, we're going to need -qmkshrobj.
>>>I'm going to work on adding it this weekend. By default, -qmkshrobj
>>>will export all symbols. We create our own export list for xlc C
>>>shared libraries. Do we want to create our own export list for xlc C++
>>>shared libraries as well?
>>>
>>>
>>>
>>>      
>>>
>>-- 
>>Jim Edwards             address@hidden
>>IBM Applications Analyst
>>NCAR SCD
>>BOULDER CO  303-497-1842 
>>
>>
>>
>>
>>_______________________________________________
>>Libtool-patches mailing list
>>address@hidden
>>http://mail.gnu.org/mailman/listinfo/libtool-patches
>>    
>>
>
>  
>


-- 
Jim Edwards             address@hidden
IBM Applications Analyst
NCAR SCD
BOULDER CO  303-497-1842 






reply via email to

[Prev in Thread] Current Thread [Next in Thread]