Roland Mainz wrote:
> "Laszlo (Laca) Peter" wrote:
>> On Tue, 2008-04-15 at 05:20 +0200, Roland Mainz wrote:
>>> Who did the libtool putback into the SFW gate ?
>> That's a strange question.
> 
> Why ?
> 
>> Does it matter?
> 
> Well, I'm interested to find someone who can guide me through my first
> putback to SFWNV and one of my two choices (the other one is
> http://defect.opensolaris.org/bz/show_bug.cgi?id=868) is to improve
> libtool performance (no ARC case required - I just want to switch from
> /usr/bin/bash to /usr/bin/ksh93 and maybe use "shcomp" - that's more
> than good enougth for the FOX bits) to help my own team a bit...

There are more pressing problems with libtool which probably need to 
be addressed, before we start worrying about libtool performance. For 
one, i doubt anyone expects it to be a showcase of HPC. Being less of 
a pain to deal with would be a small step for libtool, and a giant 
leap for humanity.

For starters, there are several Sun Studio flags in the 
-W<something>,<something> [ examples: -Wc,<something> or 
-Wu,<something>, etc ] category. These flags aren't used very often, 
but they are very useful.

libtool silently removes some of these flags (but not all), because it 
does not like them. Ideally, it should allow us to pass any compiler 
flags verbatim.

Second, there should be a libtool option to pass -norunpath when 
linking a shared object or an executable. Currently this is not 
possible without additional acrobatics.

These two enhancements would be more useful to a wider audience than 
switching libtool shells for a 5% speed improvement.

--Stefan

-- 
Stefan Teleman
Sun Microsystems, Inc.
stefan.teleman at Sun.COM


Reply via email to