> > Until recently the modules did still appear in 00modlist.long.html
> > but now they have vanished from there too.
> I started investigating for Crypt::Tea. Are you sure, it ever was
> registered in the module list? I don't think so.

No, I'm not sure; but I do think I remember checking 00modlist.long.html
after I first registered the modules back in December and feeling happy
because they were there; but I suppose the memory might be mistaken.

> I'm not saying it should not be registered, but right now, for the
> sake of debugging, I need to know if it ever was in. I cannot find
> a record of it being registered.

On Tue, 18 Dec 2001, [EMAIL PROTECTED] wrote:
> Subject: Module submission Crypt::Tea
> The following module was proposed for inclusion in the Module List:
>   modid:       Crypt::Tea
>   DSLIP:       adhfp
>   description: TEA, the Tiny Encryption Algorithm
>   userid:      PJB (Peter Billam)
>   chapterid:   14 (Security_and_Encryption)
>   ....
>   enteredby:   PJB (Peter Billam)
>   enteredon:   Tue Dec 18 20:01:51 2001 GMT
> The resulting entry would be:
> Crypt::
> ::Tea          adhfp TEA, the Tiny Encryption Algorithm         PJB
> Thanks for registering,
> The Pause Team
> PS: The following links are only valid for module list maintainers:
> Registration form with editing capabilities: ...
> Immediate (one click) registration: ...

It did say "Thanks for registering" ... Surely that implies that I
registered ?  and I'm not a module list maintainer, just an author ...

you wrote:
> new uploads are not recognized by the indexer. Now this is the
> root of all evil. In your upload PJB/Crypt-Tea-1.36.tar.gz you set
>   $VERSION = '#VERSION#';
> PAUSE had big troubles with this version number containing "#".
> ... Since then any new upload with a $VERSION that doesn't beat
> v35.86...  is not recognized as a new version.

Ahh...

Because version numbers can occur many times in many different files
(in the Module-7.89.tar.gz name, twice in each .p[ml] file (once in
the pod), in the README, in .p[ml] files in subdirectories, in other
..p[ml] files that don't get installed and are used only in the build
process, in .p[ml] files that are installed optionally, in Makefile.PL
if it uses MakeMaker and contains VERSION=>'6.78', etc etc) at that
time my build process involved inserting the version into all these
places automatically from a single authoritative source.
(#VERSION# was of course the magic string that got replaced.)

I sort of assumed that CPAN would deduce its version number from
the Module-7.89.tar.gz name (as humans do) because it is the only
uniquely-placed source for that information.  When that failed,
I thought maybe it ran Makefile.PL with a stripped-down MakeMaker
and measured what version that gave.  It didn't occur to me that
CPAN would go grepping through all the .p[ml] files because that
seemed so unstructured and ambiguous.

Can you clarify for me which occurrence of $VERSION= in which file
is taken by CPAN as authoritative ?  Is this documented anywhere ?

> The solution to the problem is to delete Crypt-Tea-1.36.tar.gz and
> re-index the newest version as soon as Crypt-Tea-1.36.tar.gz is gone
> from CPAN. You have a menu item "Reindex" on your PAUSE menu.
>
Thanks for the suggestion. I have deleted all pre-latest versions
of everything, and will wait for 72 hours and then "Reindex" :-)

Interestingly (?) the "Delete" item also offered me all the really
old versions to delete, i.e. the ones I'd already deleted long ago :-(
I wonder if those deletions worked ...

Regards,  Peter Billam

Peter Billam  http://www.pjb.com.au [EMAIL PROTECTED]  (03) 6236 9410
GPO Box 669, Hobart TAS 7001, Australia.  Original compositions made
to be played, and Arrangements of Bach, Schubert, Brahms...  Special
Offer ! Everything Free ! And, soon . . . November at the Carousel !

Reply via email to