Mail Archives: djgpp-workers/2002/02/25/19:55:04
At 05:07 PM 2/25/02 +0200, Eli Zaretskii wrote:
>On Sun, 24 Feb 2002, Peter J. Farley III wrote:
>> It's been a while, and I'm not sure how much time I can personally
>> devote to resolving this, but Rich Dawe suggested I bring this
issue
>> to the list, so here it is.
>
>Thanks for all the footwork!
You're welcome. I can't help myself sometimes -- this stuff just cries
out to be fixed. <*Sigh*> That's what I get for being a life-long
programmer -- no resistance to the urge to "fix that one last bug..."
:)
Even when I am already over-committed and under-slept.
<Snipped>
>I would like to suggest yet another alternative, that is almost
>identical to your #4. It's a 2-step dance:
>
> 1) Don't put install-info commands into the DSM files for packages
>that
> already have entries in the DIR file from the latest released
>djdev.
>
> 2) When you do put install-info commands into the DSM, use the
>--section
> option of install-info to specify the precise section where we
>want
> the entry to be placed. If necessary, use the --entry option as
>well.
Makes good sense to me. However, rather than removing the install-info
entries in DSM's, why not just leave them all in there *with* the
appropriate DJGPP section names?
The critical requirement for this to work is the answer to the
question: What happens to info/dir when you install-info an entry that
already exists into a section that already exists? I *hope* it just
replaces the "old" entry in the same place, effectively leaving the
info/dir file intact. Does it? (I'll test this myself later, just
asking to get an "official" answer.)
>The rationale for this is that we maintain the DIR file manually, and
>any
>new ported packages are normally added to it right away (well, at
least
>that's the theory ;-). So for most packages, users who install
>something
>do not need to run install-info at all, since it's all have been
done
>for them already. The only exceptions are the packages ported since
>the last djdev release; thus clause 2) above.
Understood, but leaving the install-info commands in DSM files (with
appropriate --section and/or --entry arguments) has the added advantage
of letting "new" ports get into info/dir in the right place, even
before they get "officially" added in the CVS info/dir. Handy for
porters and for adventurous souls who volunteer to test ports.
>Of course (putting on my Texinfo co-maintainer hat), if you spot an
>Info
>manual without @dircategory or @direntry, or with faulty entries,
>please
>report that to the respective package maintainer(s). But whatever
they
>do to get their act together, we in the DJGPP project will almost
>certainly use a different partition in the DIR file, so even the
fixed
>manuals will not satisfy our specific needs.
Which is the "right" thing to do, and which I may indeed do for a few
of the packages myself. But what "categories" do we recommend to the
GNU maintainer: The ones in the "standard"? I'd make up and use a very
different list than theirs.
I will send some change recommendations to the gnu-standards buglist
after doing some archive research, and see what response I get.
>> Vis-a-vis alternative #3, there *is* a recommendation in the GNU
>> Programming Standards document about how "info/dir" files should be
>> structured and sectioned.
>
>These standards don't make much sense for DJGPP users, since they
>almost
>never build the packages themselves. So it doesn't do us any good to
>comply to the standard DIR partition, especially since the bulk of
the
>GNU project has yet to catch up with these standards. As long as
>there's
>a mess out there, we had better fix it manually, like we've been
doing
>all the time.
Agreed, which is why I brought the problem here first.
---------------------------------------------------------
Peter J. Farley III (pjfarley AT dorsai DOT org)
- Raw text -