[Rd] Depends: R (>= x.y.z)
Mon, 11 Jun 2001 19:20:47 +0200
>>>>> Thomas Lumley writes:
> On Mon, 11 Jun 2001, Paul Gilbert wrote:
KH> You can do
KH> Depends: R (<= 1.2.3)
>> Yes, but the problem is that building a package now, say for 1.3.0,
>> who knows whether one should specify
>> Depends: R (< 2.0)
>> Depends: R (< 3.0)
>> And when those versions come along the maintainer may not be around to
>> make the change to the Depends line. (An if they were they would
>> probably update the package instead.)
> Yes, a < depends should almost never be correct.
Actually, a `<' never is. Only `<=' and `>=' are allowed, according to
> However, changing it to a prefix doesn't help with this. About the
> only situation when this would apply is if we decide to break source
> compatibility (eg to put threads in). Here the CRAN maintainers will
> know what the dependency is, but it may not be fixed immediately.
> We don't want the depends to mean "was only tested with". It should mean
> "will not work without". In most cases only the package maintainer can
> distinguish between a bug (which may be exposed more clearly by a new
> version of R) and a valid dependency.
We have an informal mechanism of moving old versions into version
specific ``old'' subdirs. Here's what happens. When a new release
comes along I test all packages (argh) against the new version. The
ones that do not pass R CMD check for the new version will be moved out
to the `old' subdirs, unless a new version comes along (in which case
they are stored somewhere else).
At least, this was the old way. Now that we actually have an archive of
old packages, we maybe should put them in there as well and record the
last version they were known to work with somewhere else. (Or change
the DESCRIPTION Depends info accordingly?)
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To: email@example.com