Hi Alex, Once we start accept pull-requests, we will need to view and easily find where it is, in case you need to touch or reverse, whether one or more commits.
When we need to move, the resolution of a particular ticket between fork's, or between branches disconnected, it will be very useful to be able to jump to the branch associated with it, and be assured that the contents thereof, is complete and that no commit was out. When a ticket is complicated, or reopens, or wide, if we have been committing in the main branch individual over time, we could revise the delta changes, when this is necessary. I could ramble more examples, but although this can not be imposed, it should not abridged the use of this process, since over time the Git tree is to help maintain the contributions, not to limit ourselves to a flat line as SVN and lose that advantage. The criterion would be that a ticket, a bug, an experiment, functionality, etc. should be associated with a local branch (minimum) and a public if necessary more than one developer working on it at the same time as to the branch main "Develop", should always be in a stable state, and the mixes contain individual developments, thus "Develop" is always available to open a branch of release. But as said, is primarily a visual aid and indexed for whoever has to move things, or mount a version output. Finally, as Carlos said, it's hard to sell the idea or motivation -- Jose Barragan Chief Software Architect Codeoscopic Madrid C/. Infanta Mercedes, 92. Planta 5. 505. 28020 Madrid. Tel.: +34 912 94 80 80 On Mar 27, 2013, at 6:59 PM, Frédéric THOMAS <webdoubl...@hotmail.com> wrote: >> without taking care if a branch is 1 commit or 10, since this make you gain > control of revert something if you need at some point in time > > Are you saying it is easier to revert a set of one commit + the merge commit > than only one commit ? > Because to revert one commit is as easy as 'git revert <myCommit>' > > Thanks, > -Fred > > -----Message d'origine----- From: Carlos Rovira > Sent: Wednesday, March 27, 2013 6:47 PM > To: dev@flex.apache.org > Subject: Re: [2/2] git commit: Merge branch 'FLEX-33349' into develop > > Hi Alex, > > as I said this is a matter of taste, but we are using git with huge > projects for months and when we are merging and generating releases and we > end wanting to easily see *all* branches to quickly see what we want to do, > without taking care if a branch is 1 commit or 10, since this make you gain > control of revert something if you need at some point in time. You end > having more control and people not involved in a particular development can > easily see what happened. > > Right now I think is difficult to see since we are starting in GIT, but as > GIT graph evolve we will see more complex representations and will be > difficult to see quickly the GIT history. As I pass for that *problem* > before, I already adopted the way to make always a branch although I have > only one commit, but again, I recognized that it's a matter of tastes and > it's difficult to sell the idea if you like not do it. > > > > 2013/3/27 Alex Harui <aha...@adobe.com> > >> I"m used to SVN's linear history. I like the fact that Git can show branch >> history, and like the idea that I can do multiple commits before pushing >> and >> each of those commits is clearly shown in the history. >> >> But if I don't want to show those multiple commits or I don't have multiple >> commits, my instinct would be to just make it a linear entry. I think >> Carlos is saying there is some advantage to not doing that, and I want to >> know what that advantage is. >> >> >> On 3/27/13 10:06 AM, "Jose Barragan" <jose.barra...@codeoscopic.com> >> wrote: >> >> > That's the point Alex... >> > >> > Any solve can resolve as a single commit or in a change set, and this >> > decision, under what criterial it does? >> > -- >> > Jose Barragan >> > Chief Software Architect >> > Codeoscopic Madrid >> > C/. Infanta Mercedes, 92. >> > Planta 5. 505. >> > 28020 Madrid. >> > Tel.: +34 912 94 80 80 >> > >> > On Mar 27, 2013, at 6:02 PM, Alex Harui <aha...@adobe.com> wrote: >> > >> >> At minimum, if we can't decide on one particular practice, can we get a >> >> better understanding of what problem is solved by a single commit in a >> >> branch? Then those who are trying to decide what to do can make a >> better >> >> decision. >> >> >> >> >> >> On 3/27/13 9:36 AM, "Frédéric THOMAS" <webdoubl...@hotmail.com> wrote: >> >> >> >>> Another point Carlos, >> >>> >> >>>> this is really a matter of tastes. Each one should be happy with his >> way >> >>>> of proceed >> >>> >> >>> What should I do a git wiki with the good practices if at the end >> everyone >> >>> does what he wants, I guess we can try as written in the wiki for one >> month >> >>> and if it doesn't fit we can change it (but really, I think everyone >> will be >> >>> happy with it). >> >>> >> >>> Thanks, >> >>> -Fred >> >>> >> >>> -----Message d'origine----- >> >>> From: Frédéric THOMAS >> >>> Sent: Wednesday, March 27, 2013 5:22 PM >> >>> To: dev@flex.apache.org >> >>> Subject: Re: [2/2] git commit: Merge branch 'FLEX-33349' into develop >> >>> >> >>> Yeah, for sure, it's your point of view and happily everyone's got one >> and >> >>> not everyone the same, but from what I see from the current graph on >> the >> >>> develop branch or the one gave by Jose plus the explanations you gave, >> I >> >>> have the impression to see and hear "Why should I do simple when I can >> do >> >>> complicated". >> >>> >> >>> If you look closely to the ASCII graph I wrote, you might have noticed >> there >> >>> are no crossed lines, it's really linear, one commit/jira at time, I >> really >> >>> maintain it is more readable and reflect more what the people really >> >>> develops, there's no artificial forced merge. >> >>> >> >>> Thanks, >> >>> -Fred >> >>> >> >>> -----Message d'origine----- >> >>> From: Carlos Rovira >> >>> Sent: Wednesday, March 27, 2013 4:03 PM >> >>> To: dev@flex.apache.org >> >>> Subject: Re: [2/2] git commit: Merge branch 'FLEX-33349' into develop >> >>> >> >>> IMHO this is hard to see, and even more for people outside or coming >> some >> >>> time later in time, but that's only my opinion... I think we should >>> >> >>> not >> >>> stand too much on these topics and going forward since this is really >> >>> >>> a >> >>> matter of tastes. Each one should be happy with his way of proceed >> while >> >>> making things properly since both ways are right. This is much more >> like >> >>> trying to realize how to format code, each developer will do in its >>> >> >>> own >> >>> way....put the bracket in the same line? in the next line?. I think we >> >>> should be more interested in the overall GIT workflow that ensure that >> we >> >>> all can share our contributions and make the repo safe of problems. >> >>> >> >>> >> >>> >> >>> 2013/3/27 Frédéric THOMAS <webdoubl...@hotmail.com> >> >>> >> >>>> It should have been like that even: >> >>>> >> >>>> * Merge branch ŒFLEX-33451¹ into develop [Fred] >> >>>> * FLEX-33451: Fixed .gitignore [Fred] >> >>>> * FLEX-33451: Tempora... [Fred] >> >>>> * FLEX-33451: Fix TLF [Fred] >> >>>> * FLEX-33349: Fix type error [Carlos] >> >>>> * removed copy of empty.bundles... [Justin] >> >>>> * FLEX-28946 committed patch... [Cyrill] >> >>>> * Merge branch ŒFLEX-21066¹ into develop [Carlos] >> >>>> * FLEX-21066: implemented remove item... [Carlos] >> >>>> * FLEX-21066: add removeItemError... [Carlos] >> >>>> * FLEX-21066: add removeItem to list... [Carlos] >> >>>> * Merge branch ŒFLEX-33408¹ into develop [Carlos] >> >>>> * ... >> >>>> >> >>>> I hope that will be display well. >> >>>> >> >>>> -Fred >> >>>> >> >>>> *From:* Jose Barragan <jose.barra...@codeoscopic.com> >> >>>> *Sent:* Wednesday, March 27, 2013 1:26 PM >> >>>> *To:* dev@flex.apache.org >> >>>> *Subject:* Re: [2/2] git commit: Merge branch 'FLEX-33349' into >> develop >> >>>> >> >>>> Here's how it would Develop, after merging the branches of Charles >>>> >> >>>> and >> >>>> Fred, as you can see, thereby properly appreciate the content of the >> two >> >>>> applied Tickets. >> >>>> However, Justin's commit b5da14a, is visually hidden under the branch >> of >> >>>> Cyrill. >> >>>> >> >>>> Thanks, >> >>>> -- >> >>>> *Jose Barragan* >> >>>> *Chief **Software Architect* >> >>>> Codeoscopic Madrid >> >>>> C/. Infanta Mercedes, 92. >> >>>> Planta 5. 505. >> >>>> 28020 Madrid. >> >>>> Tel.: +34 912 94 80 80 >> >>>> >> >>>> On Mar 27, 2013, at 11:13 AM, Frédéric THOMAS < >> webdoubl...@hotmail.com> >> >>>> wrote: >> >>>> >> >>>> Having the rule of one commit doesn't generated an extra merge commit >> and >> >>>> makes the log history more readable and because you know that >> corresponds >> >>>> to 1 single task/jira/file, it's still easy to find it back without >> the >> >>>> visual pollution of the extra useless merge commit. >> >>>> >> >>>> For the dictator-lieutenant model, we're not as big as linux kernel, >> maybe >> >>>> one day, in between I wouldn't like to introduce hierarchy in an >> apache >> >>>> model where it doesn't fit, anyway, look at what I said for the >> bugfix I >> >>>> just shared, "I'll do the merge at the end" which means I'll re-write >> >>>> commits in order, clean up what has to be clean if any, etc.., it's >> not >> >>>> the >> >>>> dictator-lieutenant model but a collaborative one which say as a lazy >> >>>> consensus someone will merge/ clean up at the end, as you can see at >> the >> >>>> end, no needs for a dictator model to do the same things as it is >> useless >> >>>> to use a gun to kill a fly. >> >>>> >> >>>> Thanks, >> >>>> -Fred >> >>>> >> >>>> -----Message d'origine----- From: Carlos Rovira >> >>>> Sent: Wednesday, March 27, 2013 10:24 AM >> >>>> To: dev@flex.apache.org >> >>>> Subject: Re: [2/2] git commit: Merge branch 'FLEX-33349' into develop >> >>>> >> >>>> Hi Frederic, >> >>>> >> >>>> in our experience working with GIT we saw that was extremely helpful >> to >> >>>> have visual track of each set of commits that are a single fix, >> >>>> functionality, feature or whatever, since you can always operate and >> have >> >>>> much control as branches grow and ramifies. In the particular case >>>> >> >>>> you >> >>>> point (single commit - flat history) it's a matter of tastes but >> making it >> >>>> flat for this particular case makes you lost visibility as GIT >> workflow >> >>>> evolve over time. Right now we could make it flat (at this point of >> >>>> simplicity), but I recommend not to do it since in few months we >> should >> >>>> expect more complex GIT graphs and this kind of loops will help to >>>> >> >>>> see >> >>>> others what historically happen. >> >>>> >> >>>> Apache Flex is so huge and modular and over time we should organize >> in a >> >>>> way that make us possible to plan huge releases with lots of modules, >> >>>> changes and consistence between pieces. We are right now at the very >> >>>> beginning trying to get the basic management and functionality but as >> >>>> people will understand the full potential of the tool and the kind of >> >>>> thinks we can now do, things will complicate. As we discussed in >>>> >> >>>> other >> >>>> thread some weeks ago, I suspect that we will end in an >> >>>> "dictator-lieutenant-apache" model (as we talked with Beltran the >> apache >> >>>> part is that decisions must be consolidated in this list to adopt >>>> >> >>>> such >> >>>> model), since it's what I see in other big open source projects where >> sub >> >>>> teams are organized in repositories and they commit in such >> repositories >> >>>> while the "lieutenant" assemble the final "module" (this happen in >>>> >> >>>> the >> >>>> same >> >>>> way all the way to root until reach the "dictator"). As we said, if >> we end >> >>>> in this model, the apache way will be less restricted in repo >> permissions >> >>>> and more conducted by the list where people will assign the tasks to >> >>>> >>>> a >> >>>> single person. >> >>>> >> >>>> >> >>>> >> >>>> >> >>>> >> >>>> 2013/3/27 Frédéric THOMAS <webdoubl...@hotmail.com> >> >>>> >> >>>> Hi Carlos, >> >>>> >> >>>> This merge you did make me think I didn't talk about this case on the >> wiki >> >>>> [1], so I updated it, in short, while it is good to start a branch as >> you >> >>>> did for a new jira ticket and as you may don't know the final number >> of >> >>>> commits you will have at the end, once it is the time to merge, you >> know >> >>>> the number of commits you did, if you realize you've got only one, it >> is >> >>>> better to do a 'git rebase <my_branch>' instead of a 'git merge >> --no-ff >> >>>> <my_branch>, the reason behind that is that you can avoid the extra >> merge >> >>>> commit, practically nothing change, except you will have a flat >> history >> >>>> which is what we want for only one commit and it could be >> reverse/reset >> >>>> the >> >>>> same if needed. >> >>>> >> >>>> Thanks, >> >>>> -Fred >> >>>> >> >>>> [1] https://cwiki.apache.org/**confluence/display/FLEX/Good+** >> >>>> vs+Bad+Git+usage< >> >>>> >> https://cwiki.apache.org/confluence/display/FLEX/Good+vs+Bad+Git+usage> >> >>>> >> >>>> -----Message d'origine----- From: carlosrov...@apache.org >> >>>> Sent: Wednesday, March 27, 2013 3:12 AM >> >>>> To: comm...@flex.apache.org >> >>>> Subject: [2/2] git commit: Merge branch 'FLEX-33349' into develop >> >>>> >> >>>> Merge branch 'FLEX-33349' into develop >> >>>> >> >>>> * FLEX-33349: >> >>>> Fix TypeError #1009 happening in dataProviderRefreshed() of List.as >> after >> >>>> refreshing the dataProvider of Combobox. >> >>>> >> >>>> >> >>>> Project: >> >>>> http://git-wip-us.apache.org/**repos/asf/flex-sdk/repo< >> http://git-wip-us.ap >> >>>> ac >> >>>> he.org/repos/asf/flex-sdk/repo> >> >>>> Commit: http://git-wip-us.apache.org/**repos/asf/flex-sdk/commit/** >> >>>> 39fdf7fa < >> http://git-wip-us.apache.org/repos/asf/flex-sdk/commit/39fdf7fa> >> >>>> Tree: >> >>>> http://git-wip-us.apache.org/**repos/asf/flex-sdk/tree/**39fdf7fa< >> http://gi >> >>>> t- >> >>>> wip-us.apache.org/repos/asf/flex-sdk/tree/39fdf7fa> >> >>>> Diff: >> >>>> http://git-wip-us.apache.org/**repos/asf/flex-sdk/diff/**39fdf7fa< >> http://gi >> >>>> t- >> >>>> wip-us.apache.org/repos/asf/flex-sdk/diff/39fdf7fa> >> >>>> >> >>>> Branch: refs/heads/develop >> >>>> Commit: 39fdf7fa81329fa60eb95efc374375**a901c34d3d >> >>>> Parents: 6282657 5ca083e >> >>>> Author: Carlos Rovira <carlos.rov...@gmail.com> >> >>>> Authored: Wed Mar 27 03:12:08 2013 +0100 >> >>>> Committer: Carlos Rovira <carlos.rov...@gmail.com> >> >>>> Committed: Wed Mar 27 03:12:08 2013 +0100 >> >>>> >> >>>> >> ------------------------------**------------------------------**---------- >> >>>> .../projects/spark/src/spark/**components/List.as | 23 >> >>>> +++++++++++---- >> >>>> 1 files changed, 17 insertions(+), 6 deletions(-) >> >>>> >> ------------------------------**------------------------------**---------- >> >>>> >> >>>> >> >>>> >> >>>> >> >>>> >> >>>> -- >> >>>> Carlos Rovira >> >>>> Director de Tecnología >> >>>> M: +34 607 22 60 05 >> >>>> F: +34 912 94 80 80 >> >>>> http://www.codeoscopic.com >> >>>> http://www.directwriter.es >> >>>> http://www.avant2.es >> >>>> >> >>>> >> >>>> >> >>> >> >>> >> >> >> >> -- >> >> Alex Harui >> >> Flex SDK Team >> >> Adobe Systems, Inc. >> >> http://blogs.adobe.com/aharui >> >> >> > >> >> -- >> Alex Harui >> Flex SDK Team >> Adobe Systems, Inc. >> http://blogs.adobe.com/aharui >> >> > > > -- > Carlos Rovira > Director de Tecnología > M: +34 607 22 60 05 > F: +34 912 94 80 80 > http://www.codeoscopic.com > http://www.directwriter.es > http://www.avant2.es
