Le 23/08/2015 22:00, Yuriy Tymchuk a écrit :
Thanks Thierry!
I’m not completely aware about the issue details, but maybe it makes sense to
change something in metacello itself?
No, not really, or maybe in a very complex way...
The issue is inside Monticello, in the way the Monticello GUI interacts
with each package version history, and how that interacts with the lazy
version info I added late in the Pharo 4 development cycle.
The Monticello GUI is designed to be able to display the newer/not newer
version feedback (you know, the underlined, bold, non bold package
version status you see in a repository) out of two things: first the
complete knowledge of the complete version history of all the packages
loaded in the image. And second, the names of the mcz files in the
repository. Why the names? To avoid downloading all the mcz in the
repository to get their version history... Monticello just pray and
hopes the names are right... Oh, by the way, Gofer also prays and hopes
that the names are right when you are loading something via a
configuration and a script[*].
When I added[**] lazyness to the version history in the image, I removed
that complete history Monticello relies on (and a bit of memory out of
the image). And, as a result, this has immediately created issues in the
GUI that are taken care of as they come. Some are simple, like the
gitfiletree ones, because you get all the proper version history with a
few git commands in the repository. But you can't do that with a http
repository (smalltalkhub, squeaksource) since you can't get the remote
repository version history.
Thierry
[*] Which is very fun because, at times, the names aren't right... and
the wrong package get loaded. Incomprehensible pain and confusion ensue:(
[**] I'm the second guy doing this lazy thing with Monticello.
Smalltalk/X seems to had (have?) the same thing before. Would like to
share the experience one day;)
Uko
On 23 Aug 2015, at 20:26, Thierry Goubier <[email protected]> wrote:
Hi Nicolas, Uko,
I added an optimisation to gitfiletree which avoids the unnecessary loading all
file names. Available on Pharo4 and Pharo5 (development and stable).
Thierry
Le 23/08/2015 09:08, Nicolas Anquetil a écrit :
worse than that
last week over a slow connection, I noticed that it "loaded all file
names" (from Pharo40Inbox in my case) over and over again
nicolas
On 22/08/2015 21:29, Yuriy Tymchuk wrote:
Hi,
I’ve noticed that when I try to open a gitfiletree repo, before
displaying its contents something from meta repos is loaded (seen on
screenshot). Why does it need file names from meta repos to display
contents of the repo which is present on my machine?
Uko
