GNU bug report logs - #77537
[PATCH 0/2] gnu: netcdf-cxx4: Move netcdf to the propagated-inputs

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 77537 in the body.
You can then email your comments to 77537 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Luca Cirrottola <luca.cirrottola <at> inria.fr>
To: guix-patches <at> gnu.org
Cc: Luca Cirrottola <luca.cirrottola <at> inria.fr>
Subject: [PATCH 0/2] gnu: netcdf-cxx4: Move netcdf to the propagated-inputs
Date: Fri,  4 Apr 2025 18:28:54 +0200

Hello Guix,
since "netcdf-cxx4" is a C++ interface to the C library "netcdf", I think the
latter should be a propagated input.
Similarly, since "hdf5-parallel-openmpi" includes mpi.h in the hdf5.h
header file, then "openmpi" should be a propagated input of
"hdf5-parallel-openmpi".
Let me know what you think.
Cheers,
Luca

Luca Cirrottola (2):
  gnu: hdf5-parallel-openmpi: move openmpi to the propagated-inputs.
  gnu: netcdf-cxx4: Move netcdf to the propagated-inputs.

 gnu/packages/maths.scm | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)


base-commit: 111aacf22fdc0314f41b737106ad70c1899a58e7
-- 
2.34.1

Message #8 received at 77537 <at> debbugs.gnu.org (full text, mbox):

From: Luca Cirrottola <luca.cirrottola <at> inria.fr>
To: 77537 <at> debbugs.gnu.org
Cc: Luca Cirrottola <luca.cirrottola <at> inria.fr>
Subject: [PATCH 1/2] gnu: hdf5-parallel-openmpi: move openmpi to the
 propagated-inputs.
Date: Fri,  4 Apr 2025 18:39:25 +0200

---
 gnu/packages/maths.scm | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/gnu/packages/maths.scm b/gnu/packages/maths.scm
index 6910b32dae..0e50e7a460 100644
--- a/gnu/packages/maths.scm
+++ b/gnu/packages/maths.scm
@@ -2075,9 +2075,9 @@ (define-public hdf-eos5
 (define-public hdf5-parallel-openmpi
   (package/inherit hdf5
     (name "hdf5-parallel-openmpi")
-    (inputs
+    (propagated-inputs
      `(("mpi" ,openmpi)
-       ,@(package-inputs hdf5)))
+       ,@(package-propagated-inputs hdf5)))
     (arguments
      (substitute-keyword-arguments (package-arguments hdf5)
        ((#:configure-flags flags)
-- 
2.34.1

Message #11 received at 77537 <at> debbugs.gnu.org (full text, mbox):

From: Luca Cirrottola <luca.cirrottola <at> inria.fr>
To: 77537 <at> debbugs.gnu.org
Cc: Luca Cirrottola <luca.cirrottola <at> inria.fr>
Subject: [PATCH 2/2] gnu: netcdf-cxx4: Move netcdf to the propagated-inputs.
Date: Fri,  4 Apr 2025 18:39:26 +0200

---
 gnu/packages/maths.scm | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/gnu/packages/maths.scm b/gnu/packages/maths.scm
index 0e50e7a460..ad49899d74 100644
--- a/gnu/packages/maths.scm
+++ b/gnu/packages/maths.scm
@@ -2432,7 +2432,8 @@ (define-public netcdf-cxx4
                                               (search-input-file
                                                %build-inputs
                                                "/lib/libhdf5.so")))))
-    (inputs (list netcdf hdf5))
+    (inputs (list hdf5))
+    (propagated-inputs (list netcdf))
     (home-page "https://github.com/Unidata/netcdf-cxx4")
     (synopsis "NetCDF C++ interface")
     (description
@@ -2447,7 +2448,8 @@ (define-public netcdf-cxx4-parallel-openmpi
     (synopsis "NetCDF C++ interface (with MPI support)")
     (inputs (modify-inputs (package-inputs netcdf-cxx4)
               (prepend openmpi)
-              (replace "hdf5" hdf5-parallel-openmpi)
+              (replace "hdf5" hdf5-parallel-openmpi)))
+    (propagated-inputs (modify-inputs (package-propagated-inputs netcdf-cxx4)
               (replace "netcdf" netcdf-parallel-openmpi)))))
 
 (define-public n2p2
-- 
2.34.1

Message #14 received at 77537 <at> debbugs.gnu.org (full text, mbox):

From: Andreas Enge <andreas <at> enge.fr>
To: Luca Cirrottola <luca.cirrottola <at> inria.fr>
Cc: 77537 <at> debbugs.gnu.org
Subject: Re: [PATCH 0/2] gnu: netcdf-cxx4: Move netcdf to the propagated-inputs
Date: Mon, 7 Apr 2025 12:43:28 +0200

Hello Luca,

Am Fri, Apr 04, 2025 at 06:28:54PM +0200 schrieb Luca Cirrottola:
> Similarly, since "hdf5-parallel-openmpi" includes mpi.h in the hdf5.h
> header file, then "openmpi" should be a propagated input of
> "hdf5-parallel-openmpi".

normally this would be a reason, but here I am not totally sure.
Should it not be possible to replace openmpi by mpich, say? So the users
would be required to add any MPI library, not necessarily openmpi, to
their profile. Or is this just theory and not practice? (If I were right,
then the package might as well be renamed to "hdf5-mpi".)

I am asking because propagation should be avoided as much as possible,
since it may lead to conflicts in profiles.

If we decide to propagate, we would need a longer commit message in the
GNU/Guix style, something like:
    gnu: hdf5-parallel-openmpi: Propagate openmpi.

    * gnu/packages/maths.scm (hdf5-parallel-openmpi)[inputs]: Move hdf5 from
    here...
    [propagated-inputs]: ...to here.

> since "netcdf-cxx4" is a C++ interface to the C library "netcdf", I think the
> latter should be a propagated input.

There would be no problem with the library ("ldd libnetcdf-cxx4.so"
shows that the absolute path to the libnetcdf library is there), but
the include file requires netcdf.h. So here I agree; the reason should
be given in a comment in the packages.

I have adapted the commit message and pushed this second patch.

Thanks,

Andreas

Message #17 received at 77537 <at> debbugs.gnu.org (full text, mbox):

From: Luca Cirrottola <luca.cirrottola <at> inria.fr>
To: Andreas Enge <andreas <at> enge.fr>
Cc: 77537 <at> debbugs.gnu.org
Subject: Re: [PATCH 0/2] gnu: netcdf-cxx4: Move netcdf to the propagated-inputs
Date: Mon, 7 Apr 2025 13:55:06 +0200 (CEST)

Hello Andreas,

> Should it not be possible to replace openmpi by mpich, say? So the users
> would be required to add any MPI library, not necessarily openmpi, to
> their profile. Or is this just theory and not practice? (If I were right,
> then the package might as well be renamed to "hdf5-mpi".)

I see your point and sorry I didn't anticipate it. Here we need an MPI
implementation to run, but not a specific one (and yes, it is quite normal
to juggle from one MPI library to another). So I retire my proposition about
this.

>> since "netcdf-cxx4" is a C++ interface to the C library "netcdf", I think the
>> latter should be a propagated input.
> 
> There would be no problem with the library ("ldd libnetcdf-cxx4.so"
> shows that the absolute path to the libnetcdf library is there), but
> the include file requires netcdf.h. So here I agree; the reason should
> be given in a comment in the packages.

Here it is less ambiguous, netcdf-cxx4 cannot live without netcdf and there
is no other implementation that could be used at runtime.

For the future, should I interpret a propagated input as "something that is
needed runtime, and should not be replaced by another implementation"?

> 
> I have adapted the commit message and pushed this second patch.
> 

Thanks!
Luca

Message #22 received at 77537-done <at> debbugs.gnu.org (full text, mbox):

From: Andreas Enge <andreas <at> enge.fr>
To: Luca Cirrottola <luca.cirrottola <at> inria.fr>
Cc: 77537-done <at> debbugs.gnu.org
Subject: Re: [PATCH 0/2] gnu: netcdf-cxx4: Move netcdf to the propagated-inputs
Date: Mon, 7 Apr 2025 14:47:33 +0200

Hello,

Am Mon, Apr 07, 2025 at 01:55:06PM +0200 schrieb Luca Cirrottola:
> Here it is less ambiguous, netcdf-cxx4 cannot live without netcdf and there
> is no other implementation that could be used at runtime.

yes, I agree.

> For the future, should I interpret a propagated input as "something that is
> needed runtime, and should not be replaced by another implementation"?

I think mpi is really special in this respect, I am not aware of other
software in the same category. And "needed at runtime" is not enough.
As written before, there is usually no problem with libraries for which
we set the rpath to an absolute path in the store. Well, then usually
the dependency sneaks in through include files.

Closing this issue.

Andreas

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.