Many devs contribute to the sci-electronics herd, but are not part of it. It would be good to have them officialy part of the group.
In order not to scare them, it should be made clear that their responsibility would be limited to the packages they already maintain.
This would increase cohesion, communication, and make our collective work more effective.
We need to collate, and create when necessary, test benches for each package. The goal is to provide arch testers and users with a way
of checking that an application is properly working.
These test benches can be anything from a set of written instructions to an automated script. They should be as simple as possible.
A non-specialist must be able to use them.
Let's not forget they must be maintained to work with the new versions of the packages.
Stabilize latest versions whenever possible, and get rid of the others. Portage should be kept as lean as possible.
- Keyword version 8.04 ~ppc.
- Stabilize version 8.04.
- Get rid of all versions prior to 8 which are considered buggy by upstream.
- See if building versions 8 and later from sources is a good idea (it's java). Current decision is no, but that may change.
- Complete stabilization of version 20060123 (see bug #142481).
- Stabilize version 20060906 when appropriate.
- Get rid of version 20060123.
- Write a newer ebuild (than version 20050820) with updated dependencies.
- Include gspiceui (see bug #74201)?
- Keyword version 1.0.2 ~alpha.
- Stabilize version 1.0.2.
- Get rid of older versions.
Quite some mess to fix, here. Version 20060708 was stabilized due to version 0.34 not playing nice with gcc 4.1
(if I remember correctly). Recently, version 0.35 was released by upstream, but in portage it appears older than
20060708 although it is more recent. Thus, users not masking versions 2006mmdd, as indicated in the ChangeLog, will
not see the new release.
- Wait for version 0.35 to be old enough to deserve stabilization.
- Then stabilize 0.35, but ask arch testers to be extra careful as not many users will have seen and thus emerged it.
As far as I know, though, it's clean.
- Get rid of all 2006mmdd versions.
- Later development versions will need to be called 0.35_p2006mmdd in order to avoid this mess again.
Denis should provide and document the ugly script he uses to generate the custom tarballs.
- Change the package name to ngspice. This is confusing for some users.
- Check bug #140402. These enhancements are
interesting, but make really sure they work and that they do not break anything that already exists. That wouldn't be
a first with ngspice.
- Investigate the use boehm-gc. Heavy simulations can make use of better memory management.