Re: Corrected tarballs for Emacs 23.2 and 23.3.

Re: Corrected tarballs for Emacs 23.2 and 23.3.

[Sven Joachim]

On 2011-08-02 18:05 +0200, Chong Yidong wrote:

> GNU Emacs 23.2 and 23.3 were released absent grammar files from which
> the parsers used in its CEDET component were generated.  These grammar
> files should have been included, as they constitute part of the real
> source code.  New versions of the tarballs that include these files have
> been uploaded to the GNU FTP server at
>
>   ftp://ftp.gnu.org/gnu/emacs/emacs-23.2a.tar.bz2
>   ftp://ftp.gnu.org/gnu/emacs/emacs-23.2a.tar.gz
>   ftp://ftp.gnu.org/gnu/emacs/emacs-23.3a.tar.bz2
>   ftp://ftp.gnu.org/gnu/emacs/emacs-23.3a.tar.gz
>
> All redistributors are urged to upgrade to these revised releases.
> Their md5sums are:
>
>   981cd3764b71115835801b3f4ef081c1  emacs-23.2a.tar.gz
>   3bb6d13489610c9c105625a89d689485  emacs-23.2a.tar.bz2
>   20aef9ea5b5bf8050d39f8b1e96a1c04  emacs-23.3a.tar.gz
>   f2cf8dc6f28f8ae59bc695b4ddda339c  emacs-23.3a.tar.bz2
>
> Note also that the FSF has granted permission to distribute the old
> Emacs 23.1 and 23.2 releases, notwithstanding the absence therein of the
> grammar files.

This should probably read "23.2 and 23.3" instead, since 23.1 did not
contain the CEDET files and you made no new tarball for it?

> This permission extends to versions with unrelated
> modifications, provided those versions comply with GNU GPL version 3 in
> all other respects.  See Richard Stallman's message at
>
>   http://lists.gnu.org/archive/html/emacs-devel/2011-08/msg00035.html
>
> Thank you, and sorry for any inconvenience caused.

Cheers,
       Sven

Corrected tarballs for Emacs 23.2 and 23.3

[Chong Yidong]

GNU Emacs 23.2 and 23.3 were released without some grammar files from
which the parsers used in its CEDET component were generated.  These
grammar files should have been included, as they constitute part of the
real source code.  New versions of the tarballs, marked 23.2a and 23.3a,
were uploaded, but David Engster noticed that two grammar files were
still inadvertently omitted (the grammar for the grammars themselves,
and a grammar used in the Srecode part of CEDET).

I have uploaded corrected 23.2b and 23.3b tarballs, with those grammar
files added, to the GNU FTP server at

  ftp://ftp.gnu.org/gnu/emacs/emacs-23.2b.tar.bz2
  ftp://ftp.gnu.org/gnu/emacs/emacs-23.2b.tar.gz
  ftp://ftp.gnu.org/gnu/emacs/emacs-23.3b.tar.bz2
  ftp://ftp.gnu.org/gnu/emacs/emacs-23.3b.tar.gz

All redistributors are urged to upgrade to these revised releases.
Their md5sums are:

  96a05f185ac32eccff7b06ba67808a80  emacs-23.2b.tar.gz
  f6bec0cdce31a40e879a8b7355d016fd  emacs-23.2b.tar.bz2
  f2bfd8eab7e3b34a0f7f09acc0667e71  emacs-23.3b.tar.gz
  917ce0054ef63773078a6e99b55df1ee  emacs-23.3b.tar.bz2

Note that the FSF has previously granted permission to distribute the
23.1 and 23.2 releases, notwithstanding the absence of the grammar
files, provided those versions comply with GNU GPL version 3 in all
other respects:

  http://lists.gnu.org/archive/html/info-gnu-emacs/2011-08/msg00003.html

This earlier statement covers the two missing grammar files as well.
Just to be safe, RMS will put out another statement granting permission
to distribute, which will cover the Emacs 24 pretests too.

Thanks, and sorry for any inconvenience caused.

New pretest on Monday

[Chong Yidong]

I'll make the next Emacs 24 pretest, 24.0.92, on Monday.  This pretest
will have the grammar files snafu corrected (already done on trunk).
Please plan your commits accordingly; thanks.

Re: Corrected tarballs for Emacs 23.2 and 23.3

[Ulrich Mueller]

>>>>> On Sat, 26 Nov 2011, Chong Yidong wrote:

> GNU Emacs 23.2 and 23.3 were released without some grammar files from
> which the parsers used in its CEDET component were generated.  These
> grammar files should have been included, as they constitute part of the
> real source code.  New versions of the tarballs, marked 23.2a and 23.3a,
> were uploaded, but David Engster noticed that two grammar files were
> still inadvertently omitted (the grammar for the grammars themselves,
> and a grammar used in the Srecode part of CEDET).

> I have uploaded corrected 23.2b and 23.3b tarballs, with those grammar
> files added, to the GNU FTP server at
> [...]

For some reason, lisp/cedet/{ede,semantic,srecode}/loaddefs.el are
being regenerated when building Emacs from the 23.3b tarball. The new
files have about twice the length of the ones that are distributed.
For example:

/var/tmp/portage/app-editors/emacs-23.3-r3/work/emacs-23.3 $ ls -l lisp/cedet/semantic/loaddefs.el*
-rw-r--r-- 1 ulm users 112595 Nov 26 11:01 lisp/cedet/semantic/loaddefs.el
-rw-r--r-- 1 ulm users  56415 Nov 26 04:20 lisp/cedet/semantic/loaddefs.el~

Looking into the files, I find that (except for the files' header and
footer) everything occurs twice.

This did not happen with 23.3a:

/var/tmp/portage/app-editors/emacs-23.3-r2/work/emacs-23.3 $ ls -l lisp/cedet/semantic/loaddefs.el*
-rw-r--r-- 1 ulm users 56415 Mar  8  2011 lisp/cedet/semantic/loaddefs.el

> All redistributors are urged to upgrade to these revised releases.

Ulrich

Re: Corrected tarballs for Emacs 23.2 and 23.3

[Ulrich Mueller]

>>>>> On Sat, 26 Nov 2011, Ulrich Mueller wrote:

>>>>> On Sat, 26 Nov 2011, Chong Yidong wrote:
>> I have uploaded corrected 23.2b and 23.3b tarballs, with those
>> grammar files added, to the GNU FTP server at [...]

> For some reason, lisp/cedet/{ede,semantic,srecode}/loaddefs.el
> are being regenerated when building Emacs from the 23.3b tarball.
> The new files have about twice the length of the ones that are
> distributed.

Hm, there was no answer.

Looks like the rebuild is triggered because the timestamps of all
files in the 23.3b tarball have changed. (But I guess there must also
be something wrong with the make rules, because the loaddefs files
resulting from such a rebuild should be the same as before.)

>> All redistributors are urged to upgrade to these revised releases.

Gentoo will continue distributing the 23.3a tarball, because 23.3b
would cause increased build times for our users and changes to
unrelated installed files. The latter would also imply that all eleven
of our architecture teams would have to retest the new version before
it could be moved to our stable tree.

May I assume that RMS's statement in
<http://lists.gnu.org/archive/html/info-gnu-emacs/2011-08/msg00003.html>
covers 23.3a, too? The wording there says "old Emacs 23.2 and 23.3
releases" which seems to be general enough.

Ulrich

Re: Corrected tarballs for Emacs 23.2 and 23.3

[Glenn Morris]

Ulrich Mueller wrote:

>> For some reason, lisp/cedet/{ede,semantic,srecode}/loaddefs.el
>> are being regenerated when building Emacs from the 23.3b tarball.
>> The new files have about twice the length of the ones that are
>> distributed.
[...]
> Looks like the rebuild is triggered because the timestamps of all
> files in the 23.3b tarball have changed. (But I guess there must also
> be something wrong with the make rules, because the loaddefs files
> resulting from such a rebuild should be the same as before.)

It's a bug in Emacs-23 that is not present in the current trunk.
To reproduce in emacs-23 branch:

cd lisp
make autoloads
touch cedet/semantic/decorate/mode.el
make autoloads

Now there is a duplicate entry for mode.el in cedet/semantic/loaddefs.el.
I guess it is a bug in autoload-find-destination or
autoload-file-load-name that has since been fixed.

> Gentoo will continue distributing the 23.3a tarball, because 23.3b
> would cause increased build times for our users and changes to
> unrelated installed files.

The timestamps in 23.3b are indeed messed up. I was going to suggest
some kind of fix involving touching various files, but frankly it is
probably a whole lot simpler to simply start from the 23.3a tarfile, and
just patch the etc/grammars directory with that from 23.3b, if you care
about having those files. Or distribute a separate package
emacs23-something that just adds them to an existing installation. Then
zero extra build time! :)

Re: Corrected tarballs for Emacs 23.2 and 23.3

[Ulrich Mueller]

>>>>> On Tue, 06 Dec 2011, Glenn Morris wrote:

>> Gentoo will continue distributing the 23.3a tarball, because 23.3b
>> would cause increased build times for our users and changes to
>> unrelated installed files.

> The timestamps in 23.3b are indeed messed up. I was going to suggest
> some kind of fix involving touching various files, but frankly it is
> probably a whole lot simpler to simply start from the 23.3a tarfile,
> and just patch the etc/grammars directory with that from 23.3b, if
> you care about having those files.

Indeed, we could add the diff between 23.3a and 23.3b (about 65 kB
uncompressed) to our patchball. We'd still have to distribute the
problematic 23.3a tarball though, and mirror it ourselves, because
it's gone from GNU mirrors.

It would be too much to hope for a 23.3c tarball with timestamps
fixed, I suppose?

> Or distribute a separate package emacs23-something that just adds
> them to an existing installation. Then zero extra build time! :)

Yes, and maximum confusion for users as an added bonus. ;-)

Ulrich

Re: Corrected tarballs for Emacs 23.2 and 23.3

[Chong Yidong]

Ulrich Mueller <ulm@gentoo.org> writes:

> Indeed, we could add the diff between 23.3a and 23.3b (about 65 kB
> uncompressed) to our patchball. We'd still have to distribute the
> problematic 23.3a tarball though, and mirror it ourselves, because
> it's gone from GNU mirrors.
>
> It would be too much to hope for a 23.3c tarball with timestamps
> fixed, I suppose?

Hmm, I'm not sure how the timestamps got messed up.  Otherwise 23.xb
should have no changes from 23.xa apart from the etc/grammars directory.
Sorry about that.

But I don't think the problem of increased compilation time is severe
enough to warrant making new tarballs, unless there is a demonstrable
bug in usage of the compiled Emacs.

Re: Corrected tarballs for Emacs 23.2 and 23.3

[Ulrich Mueller]

>>>>> On Fri, 09 Dec 2011, Chong Yidong wrote:

> Ulrich Mueller <ulm@gentoo.org> writes:
>> Indeed, we could add the diff between 23.3a and 23.3b (about 65 kB
>> uncompressed) to our patchball. We'd still have to distribute the
>> problematic 23.3a tarball though, and mirror it ourselves, because
>> it's gone from GNU mirrors.
>> 
>> It would be too much to hope for a 23.3c tarball with timestamps
>> fixed, I suppose?

> Hmm, I'm not sure how the timestamps got messed up. Otherwise 23.xb
> should have no changes from 23.xa apart from the etc/grammars
> directory. Sorry about that.

Apart from the timestamp, there are no other changes in the tarball
outside of the etc/grammars dir. But as noted before, some unrelated
installed files change because of the rebuild.

> But I don't think the problem of increased compilation time is
> severe enough to warrant making new tarballs, unless there is a
> demonstrable bug in usage of the compiled Emacs.

The compile time is not the problem, but the rebuild that results in
loaddefs files with duplicate entries.

And for us as a distro it's the other way around: I cannot add 23.3b
to our stable tree because of those changes in the installed image.
We must go through a complete cycle of testing, because we cannot be
sure that the changes don't cause any subtle bugs (the more since
I didn't see any upstream pretest for the 23.3b release).

Ulrich

Re: Corrected tarballs for Emacs 23.2 and 23.3

[Richard Stallman]

    But I don't think the problem of increased compilation time is severe
    enough to warrant making new tarballs, unless there is a demonstrable
    bug in usage of the compiled Emacs.

I think it justifies making new tarballs.  Slow compilation is a
pain in the neck.

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use free telephony http://directory.fsf.org/category/tel/