Re: Build failures on kernel version changes

[Nicolas Joly <njoly%pasteur.fr@localhost>]

On Mon, Jan 16, 2012 at 12:49:05AM +0100, Martin Husemann wrote:
> On Sun, Jan 15, 2012 at 09:40:13PM +0100, Nicolas Joly wrote:
> > That was my first thought, given the low number of time tools builds
> > trigger configure reruns.
> 
> Stupid question: would removing the .cache files after each complete tools
> build be any worse or better than forcing them to /dev/null?

It should be the mostly the same, except for the cache file creation
cost.

> ISTR lots of cached results even in non update runs (if/when a tool
> build triggers), but i may misinterpret the output.

Right. configure do use cached values for a first configure run. But
they all came from shell memory ac_cv_* variables that were just
computed a few lines before. At that time, the cache file does not
exist yet. Cache variables are only dumped at the end of the configure
run when substituting or generating other files (Makefile,
config.status, ...).

-- 
Nicolas Joly

Projects and Developments in Bioinformatics
Institut Pasteur, Paris.