Hi all,
I ran the snippet below in a freshly downloaded trunk image. It found a couple of differences: - Monticello is dirty because of the Mock category - FlexibleVocabularies is dirty because its empty -Info category was removed - Collections is dirty because there is an unpackaged method, WeakRegistry class>>migrateOldRegistries These discrepancies should be corrected before the release, but does anybody have an idea why this happened? MCWorkingCopy allManagers do: [:wc | wc ancestors size = 1 ifTrue: [ | parent changes | parent := wc repositoryGroup versionWithInfo: wc ancestors first. changes := wc package snapshot patchRelativeToBase: parent snapshot. wc modified: changes isEmpty not]] displayingProgress: 'Checking packages'. - Bert - |
On Wed, Jul 13, 2011 at 7:57 AM, Bert Freudenberg <[hidden email]> wrote:
> Hi all, > > I ran the snippet below in a freshly downloaded trunk image. It found a couple of differences: > > - Monticello is dirty because of the Mock category > - FlexibleVocabularies is dirty because its empty -Info category was removed > - Collections is dirty because there is an unpackaged method, WeakRegistry class>>migrateOldRegistries > > These discrepancies should be corrected before the release, but does anybody have an idea why this happened? Well, the first one is easy - running Monticello's test suite will dirty the package. This is a hassle, but I haven't found a way to prevent it that's not even more of a hassle. If you try to look at the changes, it'll tell you there aren't any and mark the package clean. Colin |
On 14.07.2011, at 19:11, Colin Putney wrote: > On Wed, Jul 13, 2011 at 7:57 AM, Bert Freudenberg <[hidden email]> wrote: >> Hi all, >> >> I ran the snippet below in a freshly downloaded trunk image. It found a couple of differences: >> >> - Monticello is dirty because of the Mock category >> - FlexibleVocabularies is dirty because its empty -Info category was removed >> - Collections is dirty because there is an unpackaged method, WeakRegistry class>>migrateOldRegistries >> >> These discrepancies should be corrected before the release, but does anybody have an idea why this happened? > > Well, the first one is easy - running Monticello's test suite will > dirty the package. This is a hassle, but I haven't found a way to > prevent it that's not even more of a hassle. If you try to look at the > changes, it'll tell you there aren't any and mark the package clean. > > Colin Nope. That's what my snippet does - look through all changes and mark the packages accordingly. This one *is* dirty. - Bert - |
On Thu, Jul 14, 2011 at 10:34 AM, Bert Freudenberg <[hidden email]> wrote:
> Nope. That's what my snippet does - look through all changes and mark the packages accordingly. This one *is* dirty. Hmm. That can happen too, but it's less common. If a test fails or a test run gets interrupted, it can leave real changes. It might also be a bug in the test suite. Colin |
In reply to this post by Bert Freudenberg
The problem with Collection appears to go back to 4.2, at least. The
method in question looks related to some of the finalization registry.. FlexibleVocabularies too - dirty in the 4.2 release (after checking changes). For Monticello, it looks like this alpha image was saved after having run the test cases. *We should not do that* because we can't trust that there aren't / won't be extranneous objects left around after running the tests which shouldn't be put into a deployment image (alpha, release or otherwise). For deploying new alpha images, how about we keep using ReleaseBuilderTrunk. We can enhance it if necessary of course, more importantly it serves as good "documentation" about how the image was deployed. On Wed, Jul 13, 2011 at 9:57 AM, Bert Freudenberg <[hidden email]> wrote: > Hi all, > > I ran the snippet below in a freshly downloaded trunk image. It found a couple of differences: > > - Monticello is dirty because of the Mock category > - FlexibleVocabularies is dirty because its empty -Info category was removed > - Collections is dirty because there is an unpackaged method, WeakRegistry class>>migrateOldRegistries > > These discrepancies should be corrected before the release, but does anybody have an idea why this happened? > > MCWorkingCopy allManagers > do: [:wc | > wc ancestors size = 1 ifTrue: [ > | parent changes | > parent := wc repositoryGroup versionWithInfo: wc ancestors first. > changes := wc package snapshot patchRelativeToBase: parent snapshot. > wc modified: changes isEmpty not]] > displayingProgress: 'Checking packages'. > > > - Bert - > > > |
In reply to this post by Colin Putney-3
On 14.07.2011, at 19:11, Colin Putney wrote:
> On Wed, Jul 13, 2011 at 7:57 AM, Bert Freudenberg <[hidden email]> wrote: >> Hi all, >> >> I ran the snippet below in a freshly downloaded trunk image. It found a couple of differences: >> >> - Monticello is dirty because of the Mock category >> - FlexibleVocabularies is dirty because its empty -Info category was removed >> - Collections is dirty because there is an unpackaged method, WeakRegistry class>>migrateOldRegistries >> >> These discrepancies should be corrected before the release, but does anybody have an idea why this happened? > > Well, the first one is easy - running Monticello's test suite will > dirty the package. This is a hassle, but I haven't found a way to > prevent it that's not even more of a hassle. If you try to look at the > changes, it'll tell you there aren't any and mark the package clean. > > Colin Nope. That's what my snippet does - look through all changes and mark the packages accordingly. This one *is* dirty. - Bert - |
In reply to this post by Colin Putney-3
Colin Putney wrote:
This is a naive question - I'm just fishing to understand more about the process...On Wed, Jul 13, 2011 at 7:57 AM, Bert Freudenberg [hidden email] wrote:Hi all, I ran the snippet below in a freshly downloaded trunk image. It found a couple of differences: - Monticello is dirty because of the Mock category - FlexibleVocabularies is dirty because its empty -Info category was removed - Collections is dirty because there is an unpackaged method, WeakRegistry class>>migrateOldRegistries These discrepancies should be corrected before the release, but does anybody have an idea why this happened?Well, the first one is easy - running Monticello's test suite will dirty the package. This is a hassle, but I haven't found a way to prevent it that's not even more of a hassle. If you try to look at the changes, it'll tell you there aren't any and mark the package clean. Colin Would you not make a copy of trunk on the server to run the Monitecello tests against, that is thrown away after the tests? For example, served up from http://testsource.squeak.com/trunk. Depending on your host operating system that could be a shallow copy-on-write copy. |
On 18.07.2011, at 17:23, Ben Coman wrote:
Look at the Monticello tests - they do not use any actual repository, but mockups. - Bert - |
Free forum by Nabble | Edit this page |