[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: SuperLU

Hi Nico

You can copy what I did for deal.ii here [1].
The updated version of debhelper is needed so that it understands ddebs. The version after 'dh_strip --ddeb-migration' is the new version number (the one you are uploading) suffixed with a '~'. 

Regards
Graham
[1] https://anonscm.debian.org/cgit/debian-science/packages/deal.ii.git/commit/?id=a3dd3b46448be9cb2107b1eb4fa0c21adee95fb7 


On 23/05/2016 14:51, Nico Schlömer wrote:

W: libsuperlu5-dbg: empty-binary-package
  Looks like debug symbols automatically went into libsuperlu5-dbgsym,
  so libsuperlu5-dbg is redundant (i.e. rename in debian/control. Or
specify -dbg as the debug package in debian/rules, e.g. with dh_strip).


If it's build automatically, can't we just get rid of the -dbg entry in
debian/control altogether?

Cheers,
Nico

On Mon, May 23, 2016 at 1:54 PM Nico Schlömer <nico.schloemer@gmail.com>
wrote:


That's dodgy how upstream handled the non-free in MATLAB/time.m.  All

they did was delete Mathwork's copyright statement. It's on their head.

Indeed. All files that are listed as exceptions from the BSD license
(except mc64ad.*) in debian/copyright now have a BSD copyright header as
well, btw. It very much looks like they are dual-licensed now. Shall we
take this for granted? I guess we could remove some complication from
debian/copyright with this.

Cheers,
Nico

On Mon, May 23, 2016 at 1:19 PM Drew Parsons <dparsons@debian.org> wrote:


Should be good to go.

That's dodgy how upstream handled the non-free in MATLAB/time.m.  All
they did was delete Mathwork's copyright statement. It's on their head.

Drew


On Mon, 2016-05-23 at 10:15 +0000, Nico Schlömer wrote:

Drew,

5.2.1 came out yesterday with some of my PRs applied. Can you import
it please? I'd then go through the remaining issues.

Cheers,
Nico

On Fri, May 20, 2016 at 9:43 PM Drew Parsons <dparsons@debian.org>
wrote:

On Fri, 2016-05-20 at 11:00 +0000, Nico Schlömer wrote:

More important is that some dfsg files are still in the git

tree.

MATLAB/{spypart.m,time.m} and DOC/*ug.pdf.


Aha yes. (Is the user guide really nonfree?)


The argument is that the pdf is the useful bit, like a binary
program.
  Just as we need the source for binary files, for the same reason
we
want the source (the latex files) for the pdf files.



Unfortunately, I have no experience with automatically creating

dfsg

tarballs, and there's a fair chance I'll do it incorrectly. I'd
appreciate if someone could take over from here.


I've tidied up the dfsg handling, listing the reject files in
debian/copyright.

The package builds fine (and petsc3.7 seems happy with it).

The last step is to check lintian,
   lintian -i superlu_5.2.0+dfsg1-1_amd64.changes

There's a handful of warnings, have a go at fixing them:

W: superlu source: dep5-copyright-license-name-not-unique
(paragraph at line 109)
  There are 2 licence short-names "permissive". Give one a distinct
  short name from the other. permissive-colamd or something.


W: superlu source: ancient-standards-version 3.9.5 (current is
3.9.8)


W: libsuperlu5-dbg: empty-binary-package
   Looks like debug symbols automatically went into libsuperlu5-
dbgsym,
   so libsuperlu5-dbg is redundant (i.e. rename in debian/control.
Or
specify -dbg as the debug package in debian/rules, e.g. with
dh_strip).

Thanks for your packaging efforts.
Drew


Cheers,
Nico

On Fri, May 20, 2016 at 12:48 PM Drew Parsons <dparsons@debian.or

g>

wrote:

On Fri, 2016-05-20 at 09:49 +0000, Nico Schlömer wrote:

I've pushed some more changes to [1] (including a patch) and

it's

now

compiling and installing alright. I guess a review would be

in

order.


Cheers,
Nico

[1] alioth:/git/debian-science/packages/superlu.git
Reply to: