MacPorts 1.9.0-rc1 regressions (was: Re: arch checking on depends_build)

[Ryan Schmidt]

On May 29, 2010, at 18:13, Joshua Root wrote:
> On 2010-5-30 09:05 , Ryan Schmidt wrote:
>> you've said there are no known regressions in 1.9.0-rc2, but this is one of the three problems that I consider regressions.
> 
> What are the other two, and why aren't there tickets for them?

Well, I already mentioned them here before, and there was some discussion, but I hadn't re-reported them because I hadn't had the time to really test that what I think happens actually happens. And of course now that I do, I can't reproduce either of them. :) The issues I was concerned with were:


1. If a port that has a platform variant is installed in MacPorts 1.8.2, then you upgrade to 1.9.0, then you install or upgrade a port that depends on that other port, you get a warning:

Warning: Skipping upgrade since $port $version_$revision >= $port $version_$revision, even though installed variants "+darwin" do not match "". Use 'upgrade --enforce-variants' to switch to the requested variants.

It seemed to me that we should skip the mismatched variants warning in cases where the mismatched variants only affect platform variants, since this is a situation practically all users of MacPorts will encounter, and we shouldn't show warnings to the user about situations that are really rather expected and not a problem.

http://lists.macosforge.org/pipermail/macports-dev/2010-April/011697.html

The discussion ended so I didn't think it had been fixed, but it looks like you fixed this with r66858?

http://trac.macports.org/changeset/66858


2. If a port is installed in MacPorts 1.8.2, then you upgrade to 1.9.0, then upgrade that port, you get a warning:

Warning: Failed to open Portfile from registry for $port @$version_$revision

Since this is a very common situation we expect all users to encounter, I didn't feel we should be issuing warnings to confuse the user.

http://lists.macosforge.org/pipermail/macports-dev/2010-April/011695.html

Jeremy L thought thought this was because the SQLite schema had changed since I started using trunk, then you said that wasn't the reason, then the discussion ended, so again I thought it hadn't been fixed, but now that I search the past commits I think you fixed this in r66854?

http://trac.macports.org/changeset/66854