Opened 11 years ago
Closed 9 years ago
#43121 closed update (fixed)
abinit 7.4.3: update to revision 1
Reported by: | cram5431@… | Owned by: | macports-tickets@… |
---|---|---|---|
Priority: | Normal | Milestone: | |
Component: | ports | Version: | 2.2.1 |
Keywords: | haspatch maintainer | Cc: | kurthindenburg (Kurt Hindenburg) |
Port: | abinit |
Description (last modified by ryandesign (Ryan Carsten Schmidt))
Abinit 7.4.3: update to revision 1:
- use of compilers and mpi portgroups
- automatic detection of selected mpi
- addition of fftw3-single dependency
- addition of no_mpi variant
Ticket for revision 0: #42017
Attachments (3)
Change History (11)
Changed 11 years ago by cram5431@…
Changed 11 years ago by cram5431@…
Attachment: | patch-bigdft-config.diff added |
---|
comment:1 Changed 11 years ago by ryandesign (Ryan Carsten Schmidt)
Description: | modified (diff) |
---|---|
Keywords: | haspatch maintainer added |
Changed 11 years ago by cram5431@…
Attachment: | abinit-7.4.3-Portfile-rev0-rev1.diff added |
---|
Portfile rev0 vs rev1: this time in unidiff version
comment:3 Changed 11 years ago by ryandesign (Ryan Carsten Schmidt)
Thanks. A couple observations:
- I see you added a "no_mpi" variant. For several years, we've been phasing out variants whose names begin with "no_", and we don't want to add any such new negatively-named variants; they should no longer be necessary. If a user doesn't want a functionality, they simply don't select the corresponding positively-named variant, or if that variant is on by default, they deselect it. MacPorts did not used to remember deselected variants during upgrades, but that bug was fixed years ago, removing the need for negatively-named variants.
- I see you changed the presentation of ui_errors by putting a newline before and after your error text. This serves to make the display of errors in this port different from the display of errors in every other port, and that does not seem desirable.
comment:4 Changed 11 years ago by cram5431@…
Thanks for the comments and coding rules.
Point 2. is easy to solve...
Point 1. (no_mpi variant) I understand the macports coding rules. But I don't find a way to obtain the following behavior for the port:
- if a mpi variant is given, take it
- if no mpi variant is given, look for a possible selected mpi (in etc/select/mpi). If there is a default mpi, take it.
- if there is no defaut mpi, take mpich.
But, I also would like to give the possibility to compile Abinit without any mpi library (i.e. in sequential mode).
I didn't find a suitable way to do that, except with a variant for the sequential mode. Any idea ?
Could a "sequential" variant fit the coding rules ? If not, I'll drop this option.
comment:6 follow-up: 7 Changed 10 years ago by kurthindenburg (Kurt Hindenburg)
The current version in macports is 7.10.2 - is any of this patch needed?
comment:7 Changed 10 years ago by cram5431@…
Replying to khindenburg@…:
The current version in macports is 7.10.2 - is any of this patch needed?
In principle, yes. I proposed for the 7.8.2 release (see #45617), but it was not included (and not in 7.10.2). With this patch (port group detection) our experience shows that it better detects gfortran / MPI; for this reason, an unofficial port is available on the website of ABINIT (https://wiki.abinit.org/doku.php?id=tips:howto_macosx) to be used if the official port does not work.
comment:8 Changed 9 years ago by kurthindenburg (Kurt Hindenburg)
Resolution: | → fixed |
---|---|
Status: | new → closed |
moving the patches/portgroup over to ticket #45617
Please attach a unified diff of the portfile so that we can see what changes are being proposed.