On Mon 08 Sep 2008 at 03:02PM, Stephen Hahn wrote:
> * Dan Price <[EMAIL PROTECTED]> [2008-09-08 21:32]:
> > On Mon 08 Sep 2008 at 02:25PM, Dan Price wrote:
> > >
> > > So this is, in my mind, a bad package version:
> > >
> > > pkg:/[EMAIL PROTECTED]
> > >
> > > Because of the use of 01.01.
> > >
> > > What should we do about this? I was about to putback a change which
> > > correctly detects these bad versions-- but doing so will break all
> > > packaging clients.
> > >
> > > Thoughts? Urgh.
> >
> > I should clarify that the bad version numbers appear to be in
> > 'depend' actions... the actual version of SUNWmkcd is 1.1.38-0.96.
> >
> > Here are the ones I have found thus far:
> >
> > depend fmri=pkg:/[EMAIL PROTECTED] type=incorporate
> > depend fmri=pkg:/[EMAIL PROTECTED] type=incorporate
> > depend fmri=pkg:/[EMAIL PROTECTED] type=incorporate
>
> Hmm. (This might be where tainting incoming FMRIs was supposed to
> help.) Is there any benefit to having a canonical/non-canonical path
> for constructing FMRIs, where the canonical path corrects the string
> output in all cases?
>
> Can I make a package that has an FMRI with a leading '0'?
When you say FMRI, I presume you mean package name? With my proposal
for package naming, yes, 0.is.the.loneliest.number would be a valid
package name. Do you want to prevent that? It seems odd to exclude
0 but include other digits.
You can also have a version with a leading 0.
[EMAIL PROTECTED] -> valid
[EMAIL PROTECTED] -> invalid
[EMAIL PROTECTED] -> invalid
[EMAIL PROTECTED] -> invalid
After talking to Danek, I've written some code into the dependency
action which detects and corrects the brokenness. I'm not sure how
to proceed on the disto-import side of things, so I'll follow up with
Dave Comay.
-dp
--
Daniel Price - Solaris Kernel Engineering - [EMAIL PROTECTED] - blogs.sun.com/dp
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
