Re: Renaming old jakarta-commons-xyz
Nico Kadel-Garcia <[email protected]>
| Newsgroups | gmane.linux.jpackage.general |
|---|---|
| Message-ID | <[email protected]> |
On Tue, Jul 20, 2010 at 3:22 PM, David Walluck <[email protected]> wrote: > On 07/20/2010 03:19 PM, Ralph Apel wrote: >> The discussion is about my (B)R in dsol-basic >> >> - I didn't want to require apache-commons-math, because this is still in >> devel >> - I didn't want to require jakarta-commons-math, because - although >> j-c-m is provided by a-c-m - I thought it didn't make sense to require >> a name which is intended to be obsoleted >> >> But you may be right: simply (B)R apache-commons-math in this case and >> other new packages... > > If you're moving apache-* anyway, then this makes sense. > > Otherwise, requiring jakarta-commons-math isn't so bad (even in a new > package). At least it's easier to add a Provides then to go around > changes any packages that have a BR on jakarta-*. >From a great deal of painful experience with "clever" renamng, be very cautious. If one merely "Provides:" something like jakarta-commons-math, and doesn't "Obsolete:" it as well, the old one won't be flushed when installing the new package. And iif you ever decide to switch back, chaos ensues. This happedned when RHEL "Obsoleted:" the ecj package with an eclipse-ecj package, and JPackage produced far more recent releases of ecj: they were blocked by RHEL's outdated eclipse-ecj with a casually written "Obsoletes" that didn't include a version number. Would it seem reasonable to "Obsolete:" versions of the software that predate this new release?