Re: Performance issue of libtool-2.4.4

[Richard Purdie]

On Mon, 2015-02-02 at 08:19 -0600, Bob Friesenhahn wrote:
> On Mon, 2 Feb 2015, Robert Yang wrote:
> > It seems that libtool (2.4.4) has increased the packages build time, after
> > a rough investigation, it maybe because new libtoolize needs run a m4
> > command:
> 
> Did you time 'libtoolize' and compare the timings with 2.4.4 to the 
> release you were using before?
> 
> Libtoolize is not normally considered to be part of package build time 
> since packages usually already come 'libtoolized' (using some random 
> version of libtool).

I work with Robert as part of the Yocto Project (powered by OpenEmbedded
underneath). The system can build Linux from scratch for a variety of
different architectures. As part of the process we autoreconf the
software so that any changes we make to support new architectures and so
on are applied as standard everywhere. We do patch libtool to deal with
some sysroot issues I've mentioned here in the past for example.

With 2.4.3, we saw a build time of 1 hour 12 minutes for one of our
standard test targets. With 2.4.4, this increased to 1 hour 23 minutes
and we've confirmed it is this specific change. We're in the process of
testing 2.4.5 but looking at the changes, I'm not sure this will help.

As yet we've not gotten to the point of isolating the problem to
libtoolize (that is the next step) but there is a noticeable regression
in the speed of things :(. 

Put in other terms why would libtool need 15% of the time needed to
build a whole Linux OS?! :)

Cheers,

Richard