Re: Trilinos 11.x

[Nico Schlömer]

> i seem to have successfully built trilinos-11.14.2

Nice! Does this work for 12.2.1 as well?

> the shared library obj-x86_64-linux-gnu/packages/
ThreadPool/src/libtrilinos_tpi.so.11, does not belong to any package, but
> others link to it (according to dh_shlibdeps).

I had removed a bunch of libraries which are no longer actively developed
within the Trilinos framework (e.g., CTrilinos, TPI etc.),
```
   -DTrilinos_ENABLE_ThreadPool:BOOL=OFF
```
It would almost be a disservice to the community if we shipped those I
think. I do not understand how any packages still depend on ThreadPool
though.

--Nico

On Tue, Sep 22, 2015 at 5:25 PM Felix Salfelder  wrote:

> Hi there.
>
> i seem to have successfully built trilinos-11.14.2 from the master-11
> branch. the shared library
> obj-x86_64-linux-gnu/packages/ThreadPool/src/libtrilinos_tpi.so.11,
> does not belong to any package, but others link to it (according to
> dh_shlibdeps).
>
> earlier we must have had a "tpi" package, i found an old
> libtpi_11.8.deb, from before the prefixing. it contains this library.
> now, should we again introduce a now "libtrilinos_tpi" package? i don't
> know where it went, do you, Nico?
>
> cheers
> felix
>

Re: DFSG status of petsc

[Baptiste Carvello]

Hello, just a small remark (not CC-ing -legal):

Le 19/09/2015 10:46, Drew Parsons a écrit :

> The win32 binary is irrelevant to us, so it no more violates dfsg than
> a png image file would. Change its bits and distribute the altered exe
> if you want to, the licence doesn't prohibit that. Hey, why not edit it
> into the format of a png image file?

If you ever distribute win32 binaries, make sure you check them for
viruses (unless you know for sure that upstream does so).

It happened to me once that my ClamAV spotted a virus in a DLL inside a
published tarball of multiplatform Free Software; probably an infected
computer had been used for the build. It would be embarrassing for
Debian to convey such a virus, though it most probably wouldn't harm anyone.

Baptiste

Re: Bug#704782: trilinos: new package for Trilinos 11.x

[Graham Inggs]

Hi Nico / Felix

On 19/09/2015 20:13, Felix Salfelder wrote:

On Sat, Sep 19, 2015 at 06:00:33PM +, Nico Schlömer wrote:

As I said, however, the git-archive of that commit dow _not_ coincide with
what Sandia distributes as their 12.2.1 tarball.


this is business as usual, and the reason why we import tarballs into
pristine-tar. in addition, gbp even requires a dedicated tarball import
commit for some reason.


For example, the
git-archive [1] is 212 MB large, the official tarball [2] only 110 MB. This
is because some testing files are removed from the tarball. Possibly there
are other changes as well.


master is now up2date with the 12.2.1 release (tarball). it seems to
build, but not on all machines. (i only have random systems, no proper
buildbot yet).


I don't see pristine-tar data for trilinos_12.2.1.orig.tar.bz2.

I downloaded trilinos-12.2.1-Source.tar.bz2 (md5sum 
760f14cbce482b4b9a41d1c18297b531) and tried to build against it.

I received many errors about unrepresentable changes.

I then tried unpacking the source tarball and copying over the debian 
directory from git.  The build ran for about an hour then failed while 
installing with a message about being unable to find packages/panzer.


Any suggestions?

BTW, I found a site linked to from trilinos.org that doesn't require 
login and contains old and current source tarballs:

https://trilinos.org/oldsite/download/files/

Regards
Graham

Re: Bug#704782: trilinos: new package for Trilinos 11.x

[Nico Schlömer]

> How is Nico able to build his PPA then?

I'm building with the sources from master, perhaps that's the reason why.

> I'm just comparing how the old (10.0.4.dsfg-1.1) package was built in
> Debian.  The old packaging produced 5 binary packages, your new
> packaging produces 94!
> Is this really necessary?

The structure of Trilinos is much better reflected by this many packages
than it was with 5. In many ways, Trilinos works like Boost, particularly
in that it is essentially a collection of "packages". I didn't see a
disadvantage in having many packages either. Perhaps that presents a
problem somewhere?

--Nico

On Tue, Sep 22, 2015 at 2:46 PM Graham Inggs  wrote:

> On 22/09/2015 14:18, Felix Salfelder wrote:
> > it should be there, but I had to repair the pristine-tar branch once
> > more. it's not required for the binary package build, (-b if you use
> > dpkg-buildpackage).
>
> I just use 'debuild -us -uc' which does require a source tarball.
>
> >> I downloaded trilinos-12.2.1-Source.tar.bz2 (md5sum
> >> 760f14cbce482b4b9a41d1c18297b531) and tried to build against it.
> >> I received many errors about unrepresentable changes.
> >
> > the checksum is OK. i have just pushed it once again (sorry).
>
> I see the pristine-tar data now, thanks.
>
> >> I then tried unpacking the source tarball and copying over the
> >> debian directory from git.  The build ran for about an hour then
> >> failed while installing with a message about being unable to find
> >> packages/panzer.
> >>
> >> Any suggestions?
> >
> > disable panzer (also remove from LIB_PACKAGES in d/r).
> > maybe use fakeroot debian/rules binary to continue the build.
> >
> > fwiw, i have started an attempt to build version 11. it's in the
> > "master-11" branch. 12 did not work at all, eventually because it did
> > not find libmumps headers which are not part of mumps 4.10.
>
> How is Nico able to build his PPA then?
>
> I'm just comparing how the old (10.0.4.dsfg-1.1) package was built in
> Debian.  The old packaging produced 5 binary packages, your new
> packaging produces 94!
> Is this really necessary?
>
>

Re: Bug#704782: trilinos: new package for Trilinos 11.x

[Felix Salfelder]

Hi Graham.

On Tue, Sep 22, 2015 at 01:19:52PM +0200, Graham Inggs wrote:
> I don't see pristine-tar data for trilinos_12.2.1.orig.tar.bz2.

it should be there, but I had to repair the pristine-tar branch once
more. it's not required for the binary package build, (-b if you use
dpkg-buildpackage).

> I downloaded trilinos-12.2.1-Source.tar.bz2 (md5sum
> 760f14cbce482b4b9a41d1c18297b531) and tried to build against it.
> I received many errors about unrepresentable changes.

the checksum is OK. i have just pushed it once again (sorry).

> I then tried unpacking the source tarball and copying over the
> debian directory from git.  The build ran for about an hour then
> failed while installing with a message about being unable to find
> packages/panzer.
> 
> Any suggestions?

disable panzer (also remove from LIB_PACKAGES in d/r).
maybe use fakeroot debian/rules binary to continue the build.

fwiw, i have started an attempt to build version 11. it's in the
"master-11" branch. 12 did not work at all, eventually because it did
not find libmumps headers which are not part of mumps 4.10.

i'm a bit confused about what are hard errors, and what are just checks.
i am particularly confused how you got past 'configure' ...

lets see what happens next :)

cheers
felix

Trilinos 11.x

[Felix Salfelder]

Hi there.

i seem to have successfully built trilinos-11.14.2 from the master-11
branch. the shared library
obj-x86_64-linux-gnu/packages/ThreadPool/src/libtrilinos_tpi.so.11,
does not belong to any package, but others link to it (according to
dh_shlibdeps).

earlier we must have had a "tpi" package, i found an old
libtpi_11.8.deb, from before the prefixing. it contains this library.
now, should we again introduce a now "libtrilinos_tpi" package? i don't
know where it went, do you, Nico?

cheers
felix