Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Gard Spreemann]

On Wednesday 30 March 2016 18:31:48 Gianfranco Costamagna wrote:
> Hi
> 
> >http://mentors.debian.net/debian/pool/main/l/lbfgsb/lbfgsb_3.0-2.dsc
> >addresses the standards-version and the dbg package. I'll have to work
> >on the watch file and (if needed) the build system.
> 
> ok, let me know when the  other points are addressed, and I'll grab it :)
> (please call it always -1, until it goes in unstable/new queue)

Hi again,

I believe I've addressed the issues you raised now (3.0+dfsg.1-1,
[1]). The watch file with repacking currently does *not* use
Files-Excluded, but rather Ben Finney's repack scripts [2]. I'm
guessing the Files-Excluded method is preferable, but I wasn't
thinking straight when I made the change. Please let me know if it's
preferable to have it that way.

[1] 
https://mentors.debian.net/debian/pool/main/l/lbfgsb/lbfgsb_3.0+dfsg.1-1.dsc

[2] https://wiki.debian.org/BenFinney/software/repack


Best,
 Gard

Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Gianfranco Costamagna]

Hi


>http://mentors.debian.net/debian/pool/main/l/lbfgsb/lbfgsb_3.0-2.dsc
>addresses the standards-version and the dbg package. I'll have to work
>on the watch file and (if needed) the build system.


ok, let me know when the  other points are addressed, and I'll grab it :)
(please call it always -1, until it goes in unstable/new queue)

>  gfortran -g -O2 -fstack-protector-strong -fPIC -o build/lbfgsb.o -c lbfgsb.f
>so it seems the flags are taken into account. Am I mistaken?


my bad.
sorry

G.

Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Gianfranco Costamagna]

Hi,



>I see. I was under the impression that was only to be used when files
>are excluded for copyright reasons. I repackaged the upstream tarball
>because it included binaries (compiled from the source, one presumes)
>and some metadata - not necessarily things that are problematic in a
>copyright sense.


you can add a comment saying the reason for the exclude.
If there are no copyright issues instead of dfsg you should use "ds" (Debian 
Source)
suffix, and mangle debian/watch accordingly.

>This'll be more work, so I'll have to postpone it a bit. The build
>seems a bit trivial for a whole generic system to be added, doesn't
>it?


also adding a new target to the current Makefile can be a good compromise


>I think they'll be useful, yes. I'll get around to adding that.


an autopkgtest is also really helpful. and trivial to implemement
Ghislain added it to almost every package he maintains, so I think
you can take is work as starting point.

G.

Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Gard Spreemann]

On Friday 25 March 2016 18:56:40 Gianfranco Costamagna wrote:
> something needs changes:
> - std-version= 3.9.7
> - no watch file?
> - no sane build system, why are you building the library such way?
> you seem to use just two files in your library, why everything is dropped?
> I don't think flags in rules are actually evaluated, because you don't set
> them - dbg package is useless now that we have ddbg automatic generation.
> 
> Please address/comment/fix the above, and I'll do another spin

http://mentors.debian.net/debian/pool/main/l/lbfgsb/lbfgsb_3.0-2.dsc
addresses the standards-version and the dbg package. I'll have to work
on the watch file and (if needed) the build system.

As for the flags: When I debuild, I see

  gfortran -g -O2 -fstack-protector-strong -fPIC -o build/lbfgsb.o -c lbfgsb.f

so it seems the flags are taken into account. Am I mistaken?


Thanks again for the help.

Best,
Gard

Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Gard Spreemann]

On Wednesday 30 March 2016 14:27:45 Gianfranco Costamagna wrote:
> HI,
> 
> >I didn't make one since upstream's tarball (at least for the latest
> >version) contains precompiled binaries, as well as a few files outside
> >of any directory (a little tarbomb). It is my understanding that these
> >need to be stripped out of the Debian source. Should I make a script
> >for that and have uscan run it?
> >
> >Upstream has a very slow release cycle, and it is my impression that
> >the library is more or less "done". Is a watch file still important?
> 
> Filex-Excluded copyright keyword might become handy.

I see. I was under the impression that was only to be used when files
are excluded for copyright reasons. I repackaged the upstream tarball
because it included binaries (compiled from the source, one presumes)
and some metadata - not necessarily things that are problematic in a
copyright sense.

> please consider, other people might have to understand how did you
> do the work, and redo it.

That makes a lot of sense, yes.

> >There is no build system upstream. How should I best approach this
> >issue?
> 
> add one :)

This'll be more work, so I'll have to postpone it a bit. The build
seems a bit trivial for a whole generic system to be added, doesn't
it?

> >- blas.f: We use the system libblas instead of this bundled copy.
> 
> nice indeed
> 
> >- driver*.f*: These are demonstration files for how to use the
> >
> >   library, and are therefore not compiled. Should they be installed
> >   as example source files somewhere?
> 
> a package-examples might be trivial to add now, but you are the maintainer
> you have to know your users' expectations. If you think an example
> packages is useful you can add it.

I think they'll be useful, yes. I'll get around to adding that.

Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Ghislain Vaillant]

On 30/03/16 15:27, Gianfranco Costamagna wrote:

- driver*.f*: These are demonstration files for how to use the
   library, and are therefore not compiled. Should they be installed
   as example source files somewhere?



a package-examples might be trivial to add now, but you are the maintainer
you have to know your users' expectations. If you think an example
packages is useful you can add it.


In my experience, example source code is quite useful when added to a
-doc package and used for package testing with autopkgtest. That way,
your test serves the double purpose of CI-testing and documenting how
to build the examples from source.

Ghis

Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Gianfranco Costamagna]

>Filex-Excluded copyright keyword might become handy.


of course s/Filex/Files here
https://wiki.debian.org/UscanEnhancements

sorry for the typo :)

G.

Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Gianfranco Costamagna]

HI,


>I didn't make one since upstream's tarball (at least for the latest
>version) contains precompiled binaries, as well as a few files outside
>of any directory (a little tarbomb). It is my understanding that these
>need to be stripped out of the Debian source. Should I make a script
>for that and have uscan run it?
>
>Upstream has a very slow release cycle, and it is my impression that
>the library is more or less "done". Is a watch file still important?


Filex-Excluded copyright keyword might become handy.

please consider, other people might have to understand how did you
do the work, and redo it.

>There is no build system upstream. How should I best approach this
>issue?


add one :)

>- blas.f: We use the system libblas instead of this bundled copy.

nice indeed

>- driver*.f*: These are demonstration files for how to use the
>   library, and are therefore not compiled. Should they be installed
>   as example source files somewhere?


a package-examples might be trivial to add now, but you are the maintainer
you have to know your users' expectations. If you think an example
packages is useful you can add it.


>- linpack.f: We use the system liblapack instead of this bundled
>   LINPACK. See also debian/patches/replace-linpack-with-lapack.patch.


nice


>That leaves only lbfgsb.f and timer.f, as you say. The former containsthe 
>entire substance of the library.

ok

cheers!


Gianfranco

Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Gard Spreemann]

On Friday 25 March 2016 18:56:40 Gianfranco Costamagna wrote:
> Hi,

Hi, and thanks for the feedback!

> something needs changes:
> - std-version= 3.9.7

Yep, I'll update that.

> - no watch file?

I didn't make one since upstream's tarball (at least for the latest
version) contains precompiled binaries, as well as a few files outside
of any directory (a little tarbomb). It is my understanding that these
need to be stripped out of the Debian source. Should I make a script
for that and have uscan run it?

Upstream has a very slow release cycle, and it is my impression that
the library is more or less "done". Is a watch file still important?

> - no sane build system, why are you building the library such way?

There is no build system upstream. How should I best approach this
issue?

> you seem to use just two files in your library, why everything is dropped?

The following source files are not used:

 - blas.f: We use the system libblas instead of this bundled copy.

 - driver*.f*: These are demonstration files for how to use the
   library, and are therefore not compiled. Should they be installed
   as example source files somewhere?

 - linpack.f: We use the system liblapack instead of this bundled
   LINPACK. See also debian/patches/replace-linpack-with-lapack.patch.

That leaves only lbfgsb.f and timer.f, as you say. The former contains
the entire substance of the library.

> I don't think flags in rules are actually evaluated, because you
> don't set them

Thank you, I'll look into that.

> - dbg package is useless now that we have ddbg automatic generation.

Will fix. Thanks.



 Best,
 Gard

Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Gianfranco Costamagna]

control: owner -1 !
control: tags -1 moreinfo

Hi,

something needs changes:
- std-version= 3.9.7
- no watch file?
- no sane build system, why are you building the library such way?
you seem to use just two files in your library, why everything is dropped?
I don't think flags in rules are actually evaluated, because you don't set them
- dbg package is useless now that we have ddbg automatic generation.

Please address/comment/fix the above, and I'll do another spin

cheers,

G.




Il Venerdì 15 Gennaio 2016 13:45, Gard Spreemann  ha 
scritto:
Package: sponsorship-requests
Severity: wishlist

Dear mentors,

I am looking for a sponsor for my package "lbfgsb"

* Package name: lbfgsb
   Version : 3.0-1
   Upstream Author : Ciyou Zhu, Richard Byrd, Jorge Nocedal and Jose Luis 
Morales
* URL : http://users.iems.northwestern.edu/~nocedal/lbfgsb.html
* License : BSD-3-clause
   Section : math

It builds these binary packages:

liblbfgsb-dev - Limited-memory quasi-Newton bound-constrained
optimization (static library)

liblbfgsb0 - Limited-memory quasi-Newton bound-constrained
optimization

liblbfgsb0-dbg - Limited-memory quasi-Newton bound-constrained
optimization (debug symbols)


Note that upstream's tarball [1] contains precompiled binaries, and is
also a minor tarbomb. I have therefore repackaged it. You will find
that my orig tarball is a strict subset of upstream's.

My package includes two patches:

- replace-linpack-with-lapack.patch: The library code originally uses
   LINPACK (from an embedded copy). Since LINPACK has largely been
   superseded by LAPACK, this patch replaces calls to the former with
   equivalent calls to the latter. Specifically, dpofa is replaced by
   dpotrf, and dtrsl is replaced by dtrtrs.

- silence.patch: The library's documentation indicates that it will
   only write out messages when the iprint flag is greater than
   zero. There are two places where writing still happens
   unconditionally, which this patch fixes. A similar patch was also
   applied by the SciPy project (see their issue 3238).

I've used the patched package on and off for the past few years and
have not encountered problems.


To access further information about this package, please visit the following 
URL:

http://mentors.debian.net/package/lbfgsb

Alternatively, one can download the package with dget using this command:

dget -x http://mentors.debian.net/debian/pool/main/l/lbfgsb/lbfgsb_3.0-1.dsc


More information about lbfgsb can be obtained from [2].

[1] http://users.iems.northwestern.edu/~nocedal/Software/Lbfgsb.3.0.tar.gz
[2] http://users.iems.northwestern.edu/~nocedal/lbfgsb.html

Regards,
Gard Spreemann

Bug#811073: RFS: lbfgsb/3.0-1 [ITP]

[Gard Spreemann]

Package: sponsorship-requests
Severity: wishlist

Dear mentors,

I am looking for a sponsor for my package "lbfgsb"

 * Package name: lbfgsb
   Version : 3.0-1
   Upstream Author : Ciyou Zhu, Richard Byrd, Jorge Nocedal and Jose Luis 
Morales
 * URL : http://users.iems.northwestern.edu/~nocedal/lbfgsb.html
 * License : BSD-3-clause
   Section : math

It builds these binary packages:

 liblbfgsb-dev - Limited-memory quasi-Newton bound-constrained
 optimization (static library)

 liblbfgsb0 - Limited-memory quasi-Newton bound-constrained
 optimization

 liblbfgsb0-dbg - Limited-memory quasi-Newton bound-constrained
 optimization (debug symbols)


Note that upstream's tarball [1] contains precompiled binaries, and is
also a minor tarbomb. I have therefore repackaged it. You will find
that my orig tarball is a strict subset of upstream's.

My package includes two patches:

 - replace-linpack-with-lapack.patch: The library code originally uses
   LINPACK (from an embedded copy). Since LINPACK has largely been
   superseded by LAPACK, this patch replaces calls to the former with
   equivalent calls to the latter. Specifically, dpofa is replaced by
   dpotrf, and dtrsl is replaced by dtrtrs.

 - silence.patch: The library's documentation indicates that it will
   only write out messages when the iprint flag is greater than
   zero. There are two places where writing still happens
   unconditionally, which this patch fixes. A similar patch was also
   applied by the SciPy project (see their issue 3238).

I've used the patched package on and off for the past few years and
have not encountered problems.


To access further information about this package, please visit the following 
URL:

 http://mentors.debian.net/package/lbfgsb

Alternatively, one can download the package with dget using this command:

 dget -x http://mentors.debian.net/debian/pool/main/l/lbfgsb/lbfgsb_3.0-1.dsc


More information about lbfgsb can be obtained from [2].

[1] http://users.iems.northwestern.edu/~nocedal/Software/Lbfgsb.3.0.tar.gz
[2] http://users.iems.northwestern.edu/~nocedal/lbfgsb.html

Regards,
 Gard Spreemann