[OpenAFS-devel] Issue building 1.5.77 rpms for i386

[Phillip Moore]

--00163630f6b728997604921db8ac
Content-Type: text/plain; charset=ISO-8859-1

OK, that makes perfect sense.  Thanks for the history lesson :-P

Since I always prefer to build everything myself, I'm just now looking into
your mockbuild.pl script, since I am also profoundly lazy and want to
automate all the stuff I just spent the week hacking on by hand.

Can I expect that to work?  I found a few of your older blog posts about it,
but I also know that a LOT has happened lately with automated builds for
OpenAFS, and I want to make sure I'm not spending time on something that's
been abandoned.

It certainly looks like what I need.  I will have a list of kernels I care
about, and I will want to be able to quickly rebuild a complete set of
binary kmod-openafs-* rpms based on whatever release of patched source tree
is interesting.

On Fri, Oct 8, 2010 at 12:26 PM, Simon Wilkinson <sxw@inf.ed.ac.uk> wrote:

>
> On 8 Oct 2010, at 11:37, Phillip Moore wrote:
>
> > No, the problem is that the openafs.spec file behaves a little wierd for
> the i386 and i686 targets.  I confess to not really understand the practical
> difference here.
>
> Historically, all RedHat operating systems built userspace binaries for the
> i386 architecture, and kernel modules for i586 and i686. Due to the way that
> RPM worked, that would require two passes, one for the i386 target and the
> other for i686. As it was most likely that if you were building for i386 you
> wanted the user space, and for i686 that you wanted kernel modules, the spec
> file was wired that way.
>
> This is all changing, though. Current Fedoras build all packages for i686
> (in the same way as happens for x86_64), so we should probably change the
> spec file to not have this default behaviour. All of the OpenAFS builds come
> through the build system, which overrides the spec file defaults, which is
> why we haven't noticed these problems.
>
> Patches are, as always, welcome!
>
> S.
>
>

--00163630f6b728997604921db8ac
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

<div><br></div>OK, that makes perfect sense. =A0Thanks for the history less=
on :-P<br><br><div>Since I always prefer to build everything myself, I&#39;=
m just now looking into your <a href=3D"http://mockbuild.pl">mockbuild.pl</=
a> script, since I am also profoundly lazy and want to automate all the stu=
ff I just spent the week hacking on by hand.<br>
<div><br></div><div>Can I expect that to work? =A0I found a few of your old=
er blog posts about it, but I also know that a LOT has happened lately with=
 automated builds for OpenAFS, and I want to make sure I&#39;m not spending=
 time on something that&#39;s been abandoned.</div>
<div><br></div><div>It certainly looks like what I need. =A0I will have a l=
ist of kernels I care about, and I will want to be able to quickly rebuild =
a complete set of binary kmod-openafs-* rpms based on whatever release of p=
atched source tree is interesting. =A0</div>
<div><br><div class=3D"gmail_quote">On Fri, Oct 8, 2010 at 12:26 PM, Simon =
Wilkinson <span dir=3D"ltr">&lt;<a href=3D"mailto:sxw@inf.ed.ac.uk">sxw@inf=
.ed.ac.uk</a>&gt;</span> wrote:<br><blockquote class=3D"gmail_quote" style=
=3D"margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class=3D"im"><br>
On 8 Oct 2010, at 11:37, Phillip Moore wrote:<br>
<br>
&gt; No, the problem is that the openafs.spec file behaves a little wierd f=
or the i386 and i686 targets. =A0I confess to not really understand the pra=
ctical difference here.<br>
<br>
</div>Historically, all RedHat operating systems built userspace binaries f=
or the i386 architecture, and kernel modules for i586 and i686. Due to the =
way that RPM worked, that would require two passes, one for the i386 target=
 and the other for i686. As it was most likely that if you were building fo=
r i386 you wanted the user space, and for i686 that you wanted kernel modul=
es, the spec file was wired that way.<br>

<br>
This is all changing, though. Current Fedoras build all packages for i686 (=
in the same way as happens for x86_64), so we should probably change the sp=
ec file to not have this default behaviour. All of the OpenAFS builds come =
through the build system, which overrides the spec file defaults, which is =
why we haven&#39;t noticed these problems.<br>

<br>
Patches are, as always, welcome!<br>
<font color=3D"#888888"><br>
S.<br>
<br>
</font></blockquote></div><br></div></div>

--00163630f6b728997604921db8ac--