ICA/Farce

Gilles Espinasse g.esp at free.fr
Mon Oct 27 15:55:28 PDT 2008


----- Original Message ----- 
From: "Ken Moffat" <ken at linuxfromscratch.org>
To: "LFS Developers Mailinglist" <lfs-dev at linuxfromscratch.org>
Sent: Monday, October 27, 2008 11:06 PM
Subject: Re: ICA/Farce


> On Mon, Oct 27, 2008 at 06:36:07PM +0100, Gilles Espinasse wrote:
> > Selon Ken Moffat <ken at linuxfromscratch.org>:
> >
> > ...
> > >
> > >  Archaic saw unresolved differences for x86 with gcc-4.1.2.
> > > Actually, looking at his results (they're at ~/archaic) they look
> > > pretty good - blkid.tab.old and locale-archive should probably now
> > > be expected to differ, and they were the only differences between
> > > his second and third builds.
> > >
> > locale-archive is a file influenced by gzip -n flag when glibc compress
locale
> > files.
> > When the locales compressed files no more include the timestamp,
locale-archive
> > checksum no more change.
> >
> > That's what I found recently.
> >
> > Gilles
>
>  Interesting.  I'm tempted to ask how you discovered this, but since
> glibc scares me to death at the best of times, I wont ;-)
>
I was lucky for this one.
I first removed the timestamp on glibc localedata/Makefile so at every
compilation the charmaps files have always same md5. When localedef run, it
include a charmaps. If charmaps md5 is the same, you have the same
locale-archive.

>  Do you have a way of either altering the build (for testing builds)
> to remove the timestamp, or do you have a way of breaking out the
> archive contents to compare them ?
>
I am doing differently to what I saw in gsbuild because I didn't know it was
there.

That's not deliberate, more because the starting goal was different and
IPCop had no packaging system.
I have a list of every files include in the .iso, and an md5 of each file is
made.
Before a release, my goal was to check that each file with a new md5 is
include in update package.

My sort of ICA is too made differently. I am building on some different
distrib and check the resulting md5 file list. The file list has to be
ordered with LC_ALL=C sort.
Some 2.4 kernel modules include the hostname, not yet checked on 2.6.

Each file that include a compilation timestamp add more noise to this
system.
That's why I made some patches to minimize the noise.
For this reason too, I am not checking Python .pyc files as they include a
timestamp by design.
I haven't find a way to hack that (and have not search a lot too).

Gilles




More information about the lfs-dev mailing list