MailScanner installer

[David Lee]

On Mon, 13 Apr 2009, Julian Field wrote:

> Right, I have just finished a good weekend's work on the installer, and all 
> the Perl modules it installs.
> Most of the Perl modules have been upgraded for starters.
>
> The installer no longer "forces" the install of any Perl module. So you can 
> upgrade Perl independently at any time without it clashing with any of 
> MailScanner's Perl modules. Hopefully this will make quite a few of you 
> happier :-)
>
> All modules install cleanly on RedHat Enterprise 4 and 5, CentOS 4 and 5, and 
> Fedora Core 9 and 10. Those are what I have tested it on and ironed out all 
> the problems on.
>
> There are several command-line options to the installer that you should know 
> about. "./install.sh --help" will print these.

Julian: This is a great improvement.  Many, many thanks.

I've just been trying it on a CentOS 5.3 (actually 5.2 upgraded to 5.3) 
with MS 4.76.14 (yesterday) and 4.76.15 (today).

It seems to be working almost flawlessly.  That is, on a repeated MS 
"./install.sh fast" almost all modules give:
    Oh good, module File-Spec version 0.82 is already installed.
    etc.

Almost.  But TimeDate is being recalcitrant.  It is consistently 
re-installing:
    Attempting to build and install perl-TimeDate-1.16-4
    Installing perl-TimeDate-1.16-4.src.rpm
    [... lots of build info ...]
    Preparing...                ##################################################
            package perl-TimeDate-1.16-5.el5.noarch (which is newer than perl-TimeDate-1.16-4.noarch) is already installed


The OS already has "perl-TimeDate-1.16-5.el5" (note the "-5"), but MS 
seems to want to try to install its own (and earlier) "-4".


If I remove that OS-derived module, and re-run MS "install.sh" twice, it 
behaves nicely: the first run installs it (perl-TimeDate-1.16-4), and the 
second run finds it and skips it (with "Oh good, ...").

The "quick fix for now" might simply be to upgrade ths MS version to 
TimeDate to "-5".  But I suspect there is a deeper structural issue at 
play, in that your re-working of the installer seems to have missed a 
detail somewhere that is affecting the TimeDate module.


Just to re-iterate: this re-working is a massive improvement!  All I'm 
reporting here is a small, suboptimal detail.

Is there something here I can assist debugging?

-- 

:  David Lee                                I.T. Service          :
:  Senior Systems Programmer                Computer Centre       :
:  UNIX Team Leader                         Durham University     :
:                                           South Road            :
:  http://www.dur.ac.uk/t.d.lee/            Durham DH1 3LE        :
:  Phone: +44 191 334 2752                  U.K.                  :