[Pharo-project] Problem with merging, changes and MC
Hi. Together with Guille we found a problem and we were able to reproduce it. The thing is this... I have my FuelTest package with version MarcusDenker.286. Then I did a merge with the last version from Fuel repo (Name: FuelTests-MarianoMartinezPeck.289). Now, if you see the package in the MC browser, it looks like: * FuelTests (MarcusDenker.286, MarianoMartinezPeck.289) So now when I search for "changes" the #closestAncestorSnapshotIn: is wrong because it is NOT taking the last version. From what we could understand, the problem is that #closestAncestorVersionFor:ifNone: sends #breadthFirstAncestorsDo: and this last methods does not SORT the versions. So...since the .286 is first and .289 is last, it answers the wrong value. Just as a test, we did a "ancestors := ancestors reversed" in #breadthFirstAncestorsDo: and the "changes" work as expected. So...what should we do? change the #breadthFirstAncestorsDo: so that it does not depend on the "sort", fix the Ëmerge" so that it sorts the versions? something else? Thanks! -- Mariano http://marianopeck.wordpress.com
On Oct 1, 2012, at 11:19 AM, Mariano Martinez Peck wrote:
Hi. Together with Guille we found a problem and we were able to reproduce it. The thing is this... I have my FuelTest package with version MarcusDenker.286. Then I did a merge with the last version from Fuel repo (Name: FuelTests-MarianoMartinezPeck.289). Now, if you see the package in the MC browser, it looks like:
* FuelTests (MarcusDenker.286, MarianoMartinezPeck.289)
So now when I search for "changes" the #closestAncestorSnapshotIn: is wrong because it is NOT taking the last version. From what we could understand, the problem is that #closestAncestorVersionFor:ifNone: sends #breadthFirstAncestorsDo: and this last methods does not SORT the versions. So...since the .286 is first and .289 is last, it answers the wrong value. Just as a test, we did a "ancestors := ancestors reversed" in #breadthFirstAncestorsDo: and the "changes" work as expected.
So...what should we do? change the #breadthFirstAncestorsDo: so that it does not depend on the "sort", fix the Ëmerge" so that it sorts the versions? something else?
I have the impression that making sure that breadthFirstAncestors start from the latest makes sense. Stef
On 2012-10-01, at 12:40, Stéphane Ducasse <stephane.ducasse@inria.fr> wrote:
On Oct 1, 2012, at 11:19 AM, Mariano Martinez Peck wrote:
Hi. Together with Guille we found a problem and we were able to reproduce it. The thing is this... I have my FuelTest package with version MarcusDenker.286. Then I did a merge with the last version from Fuel repo (Name: FuelTests-MarianoMartinezPeck.289). Now, if you see the package in the MC browser, it looks like:
* FuelTests (MarcusDenker.286, MarianoMartinezPeck.289)
So now when I search for "changes" the #closestAncestorSnapshotIn: is wrong because it is NOT taking the last version. From what we could understand, the problem is that #closestAncestorVersionFor:ifNone: sends #breadthFirstAncestorsDo: and this last methods does not SORT the versions. So...since the .286 is first and .289 is last, it answers the wrong value. Just as a test, we did a "ancestors := ancestors reversed" in #breadthFirstAncestorsDo: and the "changes" work as expected.
So...what should we do? change the #breadthFirstAncestorsDo: so that it does not depend on the "sort", fix the Ëmerge" so that it sorts the versions? something else?
I have the impression that making sure that breadthFirstAncestors start from the latest makes sense.
indeed sounds reasonable... what I always do with all the MC stuff I have no clue about, I just set up an with the changes and use it for some while :) if it works and all the MC tests pass, OK to go :P
participants (3)
-
Camillo Bruni -
Mariano Martinez Peck -
Stéphane Ducasse