svn commit: trunk/uClibc

Mike Frysinger vapier at gentoo.org
Mon Sep 26 17:39:53 UTC 2005


On Monday 26 September 2005 10:15 am, Peter S. Mazinger wrote:
> On Mon, 26 Sep 2005, Mike Frysinger wrote:
> > On Monday 26 September 2005 04:07 am, psm at uclibc.org wrote:
> > > Author: psm
> > > Date: 2005-09-26 01:07:11 -0700 (Mon, 26 Sep 2005)
> > > New Revision: 11648
> > >
> > > Log:
> > > Correct targets install_dev and install_runtime to depend on shared
> >
> > the reason i havent done this is because it re-runs make in the entire
> > tree (can take some time to recurse through the whole tree) and for many
> > targets, will re-run `ar` and `ld` ... for example, libc.a and libc.so
> > are regenerated everytime
> >
> > is this really neccessary ?
>
> no, if buildroot does not call directly install_dev (without first
> running make all), or we could create some .compiled file and check
> that instead of the dependency. Another check could be if TOPDIR/libc.so
> exists.

actually, this is a good reason for the dependency to not exist

buildroot needs to hook in and install just the header files (before the first 
cross gcc is built) ... if buildroot is doing something wrong such that the 
libs are not being built before attempting to install, then buildroot should 
be fixed

so please revert

> I do not like the idea at all, that anytime rerunning the .so/.a files are
> recreated, we should maybe do some corrections to all the Makefiles so
> that we rebuild only if some object was updated.

sure, if you feel like tackling that, g'luck ... i deemed it not worth the 
effort ... especially the overhead of watching make recurse through the whole 
directory tree is annoying

not all libs do it, just some (like libc for sure)
-mike



More information about the uClibc mailing list