[Q] Re: [xemacs-base] Add byte-compiler warning control library

[Steve Youngs]

* Stephen J Turnbull <stephen at xemacs.org> writes:

  > Steve Youngs writes:
  >> > I'm not sure that that is correct though; the package version
  >> > should presumably be used when explicitly requiring from another
  >> > package.
  >> 
  >> That sounds like crazy talk to me. :-)

  > Not really.  More like "broken by design."  Consider, packages are
  > supposed to be version-independent, but the main use of
  > package-suppress is to cater to version-specific differences in
  > location of some libraries.  So a package wants version-independent
  > code (unless it knows what it's doing).

If a package knows what it is doing in that it needs a specific
version of some lib it should be implementing the needed functionality
itself to maintain version-independence.

  > This is not the same as replacing a broken function with a fixed one,
  > I think.  This is taking a package internal because you want to change
  > its API and semantics.  That was Ben's explicit intention.

Yes, I remember him saying along those lines.

`package-suppress' first appeared on Feb 3rd, 2005, and to the best of
my knowledge, I'm the first to say anything about it not working... in
2007.  Nobody else has said anything simply because nobody else has
noticed that `package-suppress' doesn't work.  But that isn't
surprising... 

There is currently a total of two suppressed libs in the packages,
both in xemacs-base.  Namely, `regexp-opt.el' and `easy-mmode.el'.
Looking a little closer [1]

  $ diff -bBU0 pkg-regexp-opt core-regexp-opt | wc -l
  => 0

  $ diff -bBU0 pkg-easy-mmode core-easy-mmode | \
      grep -v '\(;;;###\|@@\)' | wc -l
  => 12

No wonder nobody has noticed, the differences are non-existent to
trivial at best.

* Stephen J Turnbull <stephen at xemacs.org> writes:

  > Stephen J. Turnbull writes:

  > Oh, and there's worse: consider that unicode.el is that name of a
  > Mule-UCS library that is implemented completely independently of the
  > core's unicode.el.

OK, this is yet another perfect example of why `package-suppress'
needs to _completely_ suppress the library and not just its
autoloads.

Having `package-suppress' do _anything_ is the first step, though.

  >> My head hurts....

  > No more than it did before .... :-/

Take 2 aspirin, call me in the morning if pain persists. :-)

Footnotes: 
[1]  The grep in the 2nd command is simply to eliminate changes that
     were just insertion/deletion of autoload cookies and to get rid
     of "hunk headers" to give a better representation of amount of
     real changes.

-- 
|---<Steve Youngs>---------------<GnuPG KeyID: A94B3003>---|
|       SXEmacs - The only _______ you'll ever need.       |
|         Fill in the blank, yes, it's THAT good!          |
|------------------------------------<steve at sxemacs.org>---|
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 312 bytes
Desc: not available
Url : http://lists.xemacs.org/pipermail/xemacs-beta/attachments/20070318/0230ade9/attachment.pgp