Should CI/buildbot report rev-upgrade errors?

[Ryan Schmidt]

On Oct 11, 2021, at 14:01, Eric Gallager wrote:

> On Mon, Sep 6, 2021 at 7:59 PM Aaron Madlon-Kay wrote:
>> 
>> On Tue, Sep 7, 2021 at 4:42 AM Eric Gallager via macports-dev
>> <macports-dev at lists.macports.org> wrote:
>>> 
>>> I agree that this behavior would be helpful, however, doing so would
>>> require keeping the results of previous builds installed, whereas I
>>> think the current behavior is to do all builds from a fresh
>>> environment...
>> 
>> While the intended function of rev-upgrade would require previous
>> builds, there *are* failure cases that rev-upgrade can catch that
>> would be helpful. Case in point:
>> https://trac.macports.org/ticket/63456
>> 
>> -Aaron
> 
> Another idea that would help catch more errors with rev-upgrade
> without requiring previous builds to be kept: have the buildbots run
> port_cutleaves (or just `port uninstall unrequested`) between doing
> `port install` and `port rev-upgrade`

Buildbot workers and CI machines do not run port rev-upgrade.

Buildbot worker machines keep most ports (at least those that are depended upon by several other ports) installed (but deactivated) for performance reasons.

On buildbot worker machines there is no significance to the "requested" attribute since the build machines build all ports eventually.

Buildbot worker machines automatically uninstall ports that have been superseded by newer versions which have not yet been built, and uninstall ports which have been removed from the index.