develooper Front page | | Postings from August 2003

Re: ANNOUNCE: revitalizing "dist" (aka metaconfig)

Thread Next
H.Merijn Brand
August 11, 2003 01:42
Re: ANNOUNCE: revitalizing "dist" (aka metaconfig)
Message ID:
On Mon 11 Aug 2003 07:52, Raphael Manfredi <> wrote:
> Quoting H.Merijn Brand:
> : Either we have to take it off-list like this or move it to perl-build.
> : Anyone has a good reason not to use perl-build?
> I'm not on the perl-build list.  I only kept my perl5-porters subscription
> when I "retired" from Perl two years ago.
> How do I subscribe to perl-build?

Send an empty message to
or make the body only consist of the word "subscribe";

> : One of the thing that *I* had in mind when starting on Configure for 5.10, was
> : to add imaginary sync point to make more predictable order in Configure. A
> : sync point was to consist of an empty source file that depended on the
> : previous syncpoint and was a dependency for all the files that you wanted
> : after it. I did not do anything towards that (except thinking about it)
> : because I did not know by then that the metaconfig units were also still used
> : for previous versions of perl.
> Yes, I know what you mean.  I had the same idea.  The problem is then to
> keep track of your dependencies properly, because a "bad" dependency can
> throw off your well-balanced ordering.

Yep :)

> You know about the ?Y: laYout hints in units?  Well, assume that we make
> an empty unit End_Top.U and have it depend on all the units that have
> ?Y:TOP in their hints.  Now suppose I put up a unit with ?Y:TOP that
> requieres things like a $d_xxx variable, something that is normally computed
> after the "TOP" section.  Well, to satisfy this dependency, the ordering
> is going to be seriously shuffled!

I don't care. It's a one time reshuffling. All subsequent orders will be more

> : A second, more important, think I'm still up to start, is to enable callbacks
> : *always*, not only if the -Duse64bitall is on, but also if -U is chosen. In
> : some cases, some compilers can benifit of options if they know that some
> : things are for sure *not* true. I've heard other people cheering for this
> : thought for reasons I cannot remember. Before doing so however, all hints
> : files have to protect their cbu's to do only what they would do right now if
> : the option is indeed defined. Easy if it's for 5.10 only, but I've got no idea
> : how many snakes would escape on 5.[34568].x
> Callbacks is an idea I threw off at Andy a long time ago.  I'd have to check
> on Perl units and hints to see how it got finally implemented.

It got implemented very well :)

> Could you please send me or point me to a diff or a tar.gz where I could
> know the differences between the last dist revision "dist-3.0@70" and
> the "dist-3.0@70b" that you're using for Perl?  And also let me know where
> I can download all the private units used in Perl?

I'll send you my tree off-list. It might not be completely in sync with the
metaunits, since I've not worked on them lately, but p4 might help here

H.Merijn Brand        Amsterdam Perl Mongers (
using perl-5.6.1, 5.8.0 & 633 on HP-UX 10.20 & 11.00, AIX 4.2, AIX 4.3,
  WinNT 4, Win2K pro & WinCE 2.11.  Smoking perl CORE:
send smoke reports to:, QA:

Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About