Iceberg experience report - synchronising README change made on github
Could the Iceberg gurus try and comment on the following workflow. Please familiarise with the first four commits here... https://github.com/Traadh/cloudflareun/network from these actions... "first commit" -- on github I created it as an unitialized repo to try to push a "new repo" to it per my other post, but when that failed from the command line I did `git init; git add README; git remote ...; git push ...`) "First Pharo code" -- cloned the repo with Iceberg, added a package, synchronised, added commit comment and clicked <Commit and push onto master> "Update class comment" -- edited the class comment, then from Iceberg synchronised, added commit comment and clicked <Commit and push onto master> "Updated README" -- on github, used the built-in editor to easily preview markdown. Clicked its <Commit directly to the master branch.> PROBLEM... Now I'm at a point having trouble synchronising Iceberg to pick up that last change. 1. Right-clicked on "cloudflareun" and chose "Synchronise repository..." 2. In [Update] tab, clicked <Fetch new versions> and listed under [commits to be imported] is "Updated README". Good. 3. Then clicked on <Pull> which I expected would load that commit into the Image, and a new commit "Merge with ref/remotes/origin/master" showed up in [commit to be imported] So now I'm lost. It seems the <Pull> only operates on the disk? But why does it create a merge commit? It should have fast-forwarded. Checking from the command line, only shows the first three commits are common with the github network graph. The fourth commit is the new merge commit. $ git log commit f49c64a7751f95712acc30dab692fc7e85e0c810 Merge: c818e9a 52ab6ac Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:55:34 2018 +0800 Merge with refs/remotes/origin/master commit c818e9adebeb6be001b202aa48da009be97920eb Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:42:57 2018 +0800 Update class comment. commit 6f2df89b219c56530412252ace96bc95eb78373a Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:37:01 2018 +0800 First Pharo code commit 52ab6ac696760a60fe56af57e45f2e2ceaaae35e Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:31:28 2018 +0800 first commit Checking the remote... $ git log origin/master commit 52ab6ac696760a60fe56af57e45f2e2ceaaae35e Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:31:28 2018 +0800 first commit Whoops... not what I expected but I don't think that affected anything now. Probably it was left over from early attempts to create a "new" "cloudflareun" repo. Try again... $ git remote --v origin git@github.com:Traadh/cloudflareun.git (fetch) origin git@github.com:Traadh/cloudflareun.git (push) traadh git@github.com:Traadh/cloudflareun.git (fetch) traadh git@github.com:Traadh/cloudflareun.git (push) $ git log traadh/master commit 587bed95c4b8fb3986c9962c9206405c81c99e3d Author: Ben Coman <github.com@ben.coman.com.au> Date: Thu Feb 8 11:54:20 2018 +0800 Updated README commit c818e9adebeb6be001b202aa48da009be97920eb Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:42:57 2018 +0800 Update class comment. commit 6f2df89b219c56530412252ace96bc95eb78373a Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:37:01 2018 +0800 First Pharo code commit 52ab6ac696760a60fe56af57e45f2e2ceaaae35e Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:31:28 2018 +0800 first commit So there are the four commits matching github. So it seems that Iceberg's FETCH did the right thing updating the remote. But the disk working directory should have just fast-forwarded rather than merging. Now seeing both these commits... * Merge with ref/remotes/origin/master * Updated README in the one [New commits to be imported] seems to be mixing results from the working repo and the remote repo in a way that makes it hard to guess the consequence of any next action here. I do see a <Merge and load> option when I right-click on one of those commits. Perhaps the "load" means to load from the disk working directory into the Image(?) but its a bit hidden compared to the <Fetch> and <Pull> that are buttons. It seems dangerous to take that action right now, since I guess the Image would diverge from the disk working directory. FIXING... What I ended up doing was remove the "merge commit" from the disk working directory $ git reset --hard HEAD^ then opened a new Iceberg which for [Updates] only showed only the single commit "Updated README" fetched into the remote repo. Good. I right-clicked on that and did <Merge and load> which correctly fast-forwarded the disk working directory... $ git log commit 587bed95c4b8fb3986c9962c9206405c81c99e3d Author: Ben Coman <github.com@ben.coman.com.au> Date: Thu Feb 8 11:54:20 2018 +0800 Updated README commit c818e9adebeb6be001b202aa48da009be97920eb Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:42:57 2018 +0800 Update class comment. commit 6f2df89b219c56530412252ace96bc95eb78373a Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:37:01 2018 +0800 First Pharo code commit 52ab6ac696760a60fe56af57e45f2e2ceaaae35e Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:31:28 2018 +0800 first commit However the same "Updated README" commit still shows in [New commits to be imported] tab, but now its right-click menu item now shows only <Load> rather than <Merge and load>. So it seem to correctly identify the disk working directory is up to date but clicking on <Load> doesn't change anything. Perhaps because there is no Pharo code to load? Since only the README outside the "src" directory changed? Moving to the [Push] tab, this shows the four commits seen in the last `git log` This is confusing since I'd expect this to be empty since... $ git log > /tmp/x1 $ git log traadh/master > /tmp/x2 $ diff /tmp/x1 /tmp/x2 is empty Luckily clicking <Publish all> cleared out this list with no impact on github network graph https://github.com/Traadh/cloudflareun/network SUMMARY In the end I seem to got done what I needed. I'm enjoying Iceberg more and more as I rub off the rough edges my understanding it. Hopefully my report will help smooth things for others. My only remaining niggle is not being able to get rid of this "1 incoming", but maybe that will disappear when I've got a code update to upload. cheers -ben P.S. It might be if the Iceberg [Remotes] could have a "Log" menu item that worked like `git log`. P.P.S. I think more thought needs to be given to whether mixing updates from both the remote and disk working directory in the one pane is beneficial? As an alternative, perhaps some side-scrolling GTInspector-like panes Image <--> Working Directory <--> Remote <--> Server(github) would better represent the git model. Or maybe statically position those panes in one window with buttons on directed lines showing which way and how far a commit moves when that button is pushed.
P.P.S. I think more thought needs to be given to whether mixing updates from both the remote and disk working directory in the one pane is beneficial? As an alternative, perhaps some side-scrolling GTInspector-like panes Image <--> Working Directory <--> Remote <--> Server(github) would better represent the git model. Or maybe statically position those panes in one window with buttons on directed lines showing which way and how far a commit moves when that button is pushed.
I was saying to esteban that the UI should let the user understand where the code is so I like you
Image <--> Working Directory <--> Remote <--> Server(github)
Mine was
Image <--> Local repository <--> Remote
Stef
On 11 February 2018 at 02:28, Stephane Ducasse <stepharo.self@gmail.com> wrote:
P.P.S. I think more thought needs to be given to whether mixing updates from both the remote and disk working directory in the one pane is beneficial? As an alternative, perhaps some side-scrolling GTInspector-like panes Image <--> Working Directory <--> Remote <--> Server(github) would better represent the git model. Or maybe statically position those panes in one window with buttons on directed lines showing which way and how far a commit moves when that button is pushed.
I was saying to esteban that the UI should let the user understand where the code is so I like you
Image <--> Working Directory <--> Remote <--> Server(github)
Mine was
Image <--> Local repository <--> Remote
Stef
Actually the fourth one might be redundant. The way you find out about the fourth one is you Fetch into the third one. cheers -ben
On Thu, Feb 8, 2018 at 6:28 AM, Ben Coman <btc@openinworld.com> wrote:
Could the Iceberg gurus try and comment on the following workflow.
Please familiarise with the first four commits here... https://github.com/Traadh/cloudflareun/network
from these actions...
"first commit" -- on github I created it as an unitialized repo to try to push a "new repo" to it per my other post, but when that failed from the command line I did `git init; git add README; git remote ...; git push ...`)
"First Pharo code" -- cloned the repo with Iceberg, added a package, synchronised, added commit comment and clicked <Commit and push onto master>
"Update class comment" -- edited the class comment, then from Iceberg synchronised, added commit comment and clicked <Commit and push onto master>
"Updated README" -- on github, used the built-in editor to easily preview markdown. Clicked its <Commit directly to the master branch.>
PROBLEM...
Now I'm at a point having trouble synchronising Iceberg to pick up that last change. 1. Right-clicked on "cloudflareun" and chose "Synchronise repository..." 2. In [Update] tab, clicked <Fetch new versions> and listed under [commits to be imported] is "Updated README". Good. 3. Then clicked on <Pull> which I expected would load that commit into the Image, and a new commit "Merge with ref/remotes/origin/master" showed up in [commit to be imported]
So now I'm lost. It seems the <Pull> only operates on the disk? But why does it create a merge commit? It should have fast-forwarded.
Checking from the command line, only shows the first three commits are common with the github network graph. The fourth commit is the new merge commit. $ git log commit f49c64a7751f95712acc30dab692fc7e85e0c810 Merge: c818e9a 52ab6ac Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:55:34 2018 +0800 Merge with refs/remotes/origin/master
commit c818e9adebeb6be001b202aa48da009be97920eb Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:42:57 2018 +0800 Update class comment.
commit 6f2df89b219c56530412252ace96bc95eb78373a Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:37:01 2018 +0800 First Pharo code
commit 52ab6ac696760a60fe56af57e45f2e2ceaaae35e Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:31:28 2018 +0800 first commit
Checking the remote... $ git log origin/master commit 52ab6ac696760a60fe56af57e45f2e2ceaaae35e Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:31:28 2018 +0800 first commit
Whoops... not what I expected but I don't think that affected anything now. Probably it was left over from early attempts to create a "new" "cloudflareun" repo.
Try again... $ git remote --v origin git@github.com:Traadh/cloudflareun.git (fetch) origin git@github.com:Traadh/cloudflareun.git (push) traadh git@github.com:Traadh/cloudflareun.git (fetch) traadh git@github.com:Traadh/cloudflareun.git (push)
$ git log traadh/master commit 587bed95c4b8fb3986c9962c9206405c81c99e3d Author: Ben Coman <github.com@ben.coman.com.au> Date: Thu Feb 8 11:54:20 2018 +0800 Updated README
commit c818e9adebeb6be001b202aa48da009be97920eb Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:42:57 2018 +0800 Update class comment.
commit 6f2df89b219c56530412252ace96bc95eb78373a Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:37:01 2018 +0800 First Pharo code
commit 52ab6ac696760a60fe56af57e45f2e2ceaaae35e Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:31:28 2018 +0800 first commit
So there are the four commits matching github. So it seems that Iceberg's FETCH did the right thing updating the remote. But the disk working directory should have just fast-forwarded rather than merging.
Now seeing both these commits... * Merge with ref/remotes/origin/master * Updated README in the one [New commits to be imported] seems to be mixing results from the working repo and the remote repo in a way that makes it hard to guess the consequence of any next action here.
I do see a <Merge and load> option when I right-click on one of those commits. Perhaps the "load" means to load from the disk working directory into the Image(?) but its a bit hidden compared to the <Fetch> and <Pull> that are buttons. It seems dangerous to take that action right now, since I guess the Image would diverge from the disk working directory.
FIXING...
What I ended up doing was remove the "merge commit" from the disk working directory $ git reset --hard HEAD^
then opened a new Iceberg which for [Updates] only showed only the single commit "Updated README" fetched into the remote repo. Good. I right-clicked on that and did <Merge and load> which correctly fast-forwarded the disk working directory... $ git log commit 587bed95c4b8fb3986c9962c9206405c81c99e3d Author: Ben Coman <github.com@ben.coman.com.au> Date: Thu Feb 8 11:54:20 2018 +0800 Updated README
commit c818e9adebeb6be001b202aa48da009be97920eb Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:42:57 2018 +0800 Update class comment.
commit 6f2df89b219c56530412252ace96bc95eb78373a Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:37:01 2018 +0800 First Pharo code
commit 52ab6ac696760a60fe56af57e45f2e2ceaaae35e Author: Ben Coman <btc@openInWorld.com> Date: Thu Feb 8 11:31:28 2018 +0800 first commit
However the same "Updated README" commit still shows in [New commits to be imported] tab, but now its right-click menu item now shows only <Load> rather than <Merge and load>. So it seem to correctly identify the disk working directory is up to date but clicking on <Load> doesn't change anything. Perhaps because there is no Pharo code to load? Since only the README outside the "src" directory changed?
Moving to the [Push] tab, this shows the four commits seen in the last `git log` This is confusing since I'd expect this to be empty since... $ git log > /tmp/x1 $ git log traadh/master > /tmp/x2 $ diff /tmp/x1 /tmp/x2 is empty
Luckily clicking <Publish all> cleared out this list with no impact on github network graph https://github.com/Traadh/cloudflareun/network
SUMMARY In the end I seem to got done what I needed. I'm enjoying Iceberg more and more as I rub off the rough edges my understanding it. Hopefully my report will help smooth things for others.
My only remaining niggle is not being able to get rid of this "1 incoming", but maybe that will disappear when I've got a code update to upload.
What version of Pharo/Icerberg are you using? There is a couple of glitches/bugs in iceberg 0.6.x regarding merges. - it ignores a commit when a commit only has changes in non smalltalk code (e.g., your README file) - it creates an extra merge commit in many situations that are not necessary These two problems should be fixed in the latest development version, where we revised the merge algorithm. The pull should effectively fast-forward and the commit that only touches the README should be recognized. Now, I read a separate issue in your email: the pull button does not do the same than the "Merge and Load" button of the last commit. I think that should be fixed also, but I should check. We should sync with Esteban to see when we can integrate/merge this new version.
cheers -ben
P.S. It might be if the Iceberg [Remotes] could have a "Log" menu item that worked like `git log`.
P.P.S. I think more thought needs to be given to whether mixing updates from both the remote and disk working directory in the one pane is beneficial? As an alternative, perhaps some side-scrolling GTInspector-like panes Image <--> Working Directory <--> Remote <--> Server(github) would better represent the git model. Or maybe statically position those panes in one window with buttons on directed lines showing which way and how far a commit moves when that button is pushed.
-- Guille Polito Research Engineer Centre de Recherche en Informatique, Signal et Automatique de Lille CRIStAL - UMR 9189 French National Center for Scientific Research - *http://www.cnrs.fr <http://www.cnrs.fr>* *Web:* *http://guillep.github.io* <http://guillep.github.io> *Phone: *+33 06 52 70 66 13
participants (3)
-
Ben Coman -
Guillermo Polito -
Stephane Ducasse