Pharo + git workflow
Locked
123
123
Re: Pharo + git workflow
 
|
> On 26 Jan 2016, at 16:49, Dimitris Chloupis <[hidden email]> wrote: > > Obviously it will better fit Pharo since its made to work with smalltalk code, but that does not make it any less terrible. Just because you have one implementation of something that does not mean its good. Its just means its there and it works. > > I dont know the internal, they are not documented anyway, there are some class comments here and there but thats pretty much it. I dont even remember when was the last time monticello got an updated, I mean a serious update not just a couple of bug fixes the 2 years I have been around. > > Secondly GUI is just plain awful, Smalltalk maybe be the first or one of the first to implement guis, but those implementations never ended up to something that would be approachable and easy to use on a day to day basis, some tools suffer more from this some less, Monticello is up there with the worse design. > > Thirdly the inability of the system to version control images , audio files and other assets it defeats the central purpose of smalltalk of everything being objects with a loud "Nope !" from Monticello "Only source code is". > > So its awesome that Smalltalk , and Squeak got its own version control system, that is easy to use and Pharo inherited it. Congratulations to people behind it. But the GUI needs to go, its a bad advertisement to Pharo, and we need something that is not stuck to dark ages as you correctly pointed but for the opposite reason. Because any way you try to turn Monticello you wont find a label written "modern" on it. The label you may find on it is more like "abandonware". > > Also there like a ton of OOP languages out there using git with no major problems, the problem with smalltalk is that smalltalk is an island. > > And the problem with islands is when you end having fun with them you feel stuck since they dont provide an easy access to the outside world. > > "Git just manages blobs, text files at best. Dead text" > > Last time I checked Monticello used a format called mcz which is nothing more than a zip file containing st files, which are as you call it "dead text" files. Also I would like to remind you that git is used by the CUIS smalltalk to version control their images, I thought images are live code. > > Personally I dont see the diffirence between live and dead text. Its just text to me. The VM is the one that makes it live anyway. Yes and no. If git compares two versions, it does not understand what is in it. When Monticello compares versions, it knows about classes, methods, inheritance, it can explain diffs in the same structured (browsable) form that you wrote them. Putting Smalltalk in plain text files can be done, has been done, and was not successful, because you lose something essential by leaving the environment. The island argument is also a bit too easy: any platform can be seen as an island that needs to integrate with others, there are degrees of openness. Your Blender or Python are islands too, just like SAP, Microsoft and Oracle. Some connections/interfaces are easy, some are hard(er). It all depends on your position. When have you last looked at the source code of git ? And be honest, do you find git always easy to use ? > On Tue, Jan 26, 2016 at 5:09 PM Sven Van Caekenberghe <[hidden email]> wrote: > > > On 26 Jan 2016, at 15:59, Dimitris Chloupis <[hidden email]> wrote: > > > > To be fair my experience with pharo and git have been not always smooth either. I have the VM crashing again and again completely randomly when it was trying to pull SmaCC as dependency for my project Ephestos, had to drop SmaCC and moving my python type parsing at python side. > > > > But I find it ironic someone using Monticello, trying to equate git with dark ages, you cant get more dark ages than monticello, frankly. No offense to people who made it , its great that is in there but its full of problems and bad designs and cant even begin to be compared with Github and GIT GUI clients. Monticello is according to my personal opinion by far the worst tool of Pharo. > > No it is not. It is a version management system that understands our object and code model, a system that we control. Git just manages blobs, text files at best. Dead text. > > (This does not mean it is perfect, nor that it cannot improve, nor that we should not improve our git integration.) > > > On Tue, Jan 26, 2016 at 4:51 PM Thierry Goubier <[hidden email]> wrote: > > 2016-01-26 15:11 GMT+01:00 Sean P. DeNigris <[hidden email]>: > > NorbertHartl wrote > > > - I need to use BaselineOf instead of ConfigurationOf. Thus you cannot use > > > Versionner anymore > > > > Unfortunately. This is the biggest drag for me after switching all my > > personal projects to git (GitHub for public and BitBucket for private). I > > had gotten spoiled by Versionner and hand-editing MetaC artifacts feels like > > going back to the dark ages :/ > > > > Well, I guess copying the baselines generated by Versionner into a BaselineOf is probably a way to do it. > > > > Thierry > > > > > > > > > > ----- > > Cheers, > > Sean > > -- > > View this message in context: http://forum.world.st/Pharo-git-workflow-tp4874067p4874221.html > > Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com. > > > > |
|
"If git compares two versions, it does not understand what is in it." why it should ? its a version control system, not a refactoring tool. " When Monticello compares versions, it knows about classes, methods, inheritance, it can explain diffs in the same structured (browsable) form that you wrote them." "The island argument is also a bit too easy: any platform can be seen as an island that needs to integrate with others, there are degrees of openness. Your Blender or Python are islands too, just like SAP, Microsoft and Oracle. Some connections/interfaces are easy, some are hard(er). It all depends on your position." I completely agree in the end every API ever programm is to a degree an island. No code ever shakes in terror saying "oh my god I need to make my code 100% compatible with whatever is out there" . Its not possible. But the culture of those other languages is "lets keep things close" , "lets make coder's life" easier , "lets keep compatibility and standard/common practices" . Smalltalk does not do that because it loves experimentation, the whole smalltalk enviroment is experimentation heaven playground. Thats great because without it we dont have smalltalk but its also bad. "When have you last looked at the source code of git ?" Never ? Why should I, I dont even take a look at monticello source code. I tried to improve auto completion and my head kept spinning around and around, 10s of hours still could not figure out the code. I am not smart or knowledgable to judge Monticello on a source code basis. Maybe source code wise, core wise, Monticello is 1000 times better than Git, but my complains with Monticello is on the user level. As a smalltalker I just dont see the point of prefering Monticello over Git , or screw Git, you want mercurial ? thats fine too. I just dont see it whats the big deal with Monticello. "And be honest, do you find git always easy to use ?" I have been honest from the very start when I started pushing for git and Github. I have been crystal clear. My usage of git is super simple. I dont work in teams, I work alone, I dont even use branches, I do git pull, git add, git commit and git push. I have reverted commits a couple of times, I have reset to head sometimes because of some nasty merges and that was it. I am almost never have merge conflicts. You want me to compare my experience with monticello and StHub VS git and Github using Pharo ? Nope you dont. Do I find git always easy to use ? You know what , I am a python coder and I definetly appreciate ease of use and simplicity but lets be frank here Git was made by the same guy that made the freaking Linux kernel. He made Git not to be easy but able to manage an enormous amount of code the easy way the fastest possible way and he accomplished that goal. So do I find git easy to use ? Nope Do I care ? Nope, because I rather use something that is difficult to use but powerful and with good performance than something that is very easy to use , limited and slow. Hell , I dont even find Pharo easy to use , its power and flexibility that made me love it. On Tue, Jan 26, 2016 at 6:23 PM Sven Van Caekenberghe <[hidden email]> wrote:
|
Re: Pharo + git workflow
|
|
Sometimes I really don't understand what you see in Pharo ...
> On 26 Jan 2016, at 17:53, Dimitris Chloupis <[hidden email]> wrote: > > "If git compares two versions, it does not understand what is in it." > > why it should ? its a version control system, not a refactoring tool. > > " > When Monticello compares versions, it knows about classes, methods, inheritance, > it can explain diffs in the same structured (browsable) form that you wrote them." > > is that Monticello or is that refactoring classes and methods of pharo ? Because the way I see it is like this if you have a refactoring language, a language that can refactor its own code , that language should be able to take two diffirent pieces of code and not only show you a diff but also show you how your inheritance is affected, your classes even your live state. > > The big question here, from my side is why a version control system should do that ? Refactoring and version controlling for me at least is two very different tasks. I dont also see how a version control system that is self aware refactoring wise when already have refactoring tools in my IDE, is giving me any advantage me as a user. Most IDEs come with them anyway. > > Its not as if you will use a python IDE, or whatever language you use and the IDE will go "sorry dude I cannot tell you what git did there because it has no refactoring and I am too stupid to use my own refactoring tools". That IDE would belong to the bin. This is why we use specialized Git gui tools, they are not there just to make our screens pretty. > > "Putting Smalltalk in plain text files can be done, has been done, and was not successful, because you lose something essential by leaving the environment." > > What you lose is the live state which monticello does not store anyway, how could it ? it only stores text files, its actually far more anti-Smalltalk than git is. Because with git you can version control your pharo images or your fuel files than contain the live state or you could export the live state and live code to a binary file. Now try that with monticello. The irony is very large, Git is far more Smalltalk friendly, than Monticello is. > > "The island argument is also a bit too easy: any platform can be seen as an island that needs to integrate with others, there are degrees of openness. Your Blender or Python are islands too, just like SAP, Microsoft and Oracle. Some connections/interfaces are easy, some are hard(er). It all depends on your position." > > I completely agree in the end every API ever programm is to a degree an island. No code ever shakes in terror saying "oh my god I need to make my code 100% compatible with whatever is out there" . Its not possible. But the culture of those other languages is "lets keep things close" , "lets make coder's life" easier , "lets keep compatibility and standard/common practices" . > > Smalltalk does not do that because it loves experimentation, the whole smalltalk enviroment is experimentation heaven playground. Thats great because without it we dont have smalltalk but its also bad. > > "When have you last looked at the source code of git ?" > > Never ? Why should I, I dont even take a look at monticello source code. I tried to improve auto completion and my head kept spinning around and around, 10s of hours still could not figure out the code. I am not smart or knowledgable to judge Monticello on a source code basis. Maybe source code wise, core wise, Monticello is 1000 times better than Git, but my complains with Monticello is on the user level. As a smalltalker I just dont see the point of prefering Monticello over Git , or screw Git, you want mercurial ? thats fine too. > > I just dont see it whats the big deal with Monticello. > > "And be honest, do you find git always easy to use ?" > > I have been honest from the very start when I started pushing for git and Github. I have been crystal clear. My usage of git is super simple. > > I dont work in teams, I work alone, I dont even use branches, I do git pull, git add, git commit and git push. I have reverted commits a couple of times, I have reset to head sometimes because of some nasty merges and that was it. I am almost never have merge conflicts. > > You want me to compare my experience with monticello and StHub VS git and Github using Pharo ? Nope you dont. > > Do I find git always easy to use ? You know what , I am a python coder and I definetly appreciate ease of use and simplicity but lets be frank here Git was made by the same guy that made the freaking Linux kernel. He made Git not to be easy but able to manage an enormous amount of code the easy way the fastest possible way and he accomplished that goal. > > So do I find git easy to use ? Nope > > Do I care ? Nope, because I rather use something that is difficult to use but powerful and with good performance than something that is very easy to use , limited and slow. Hell , I dont even find Pharo easy to use , its power and flexibility that made me love it. > > > On Tue, Jan 26, 2016 at 6:23 PM Sven Van Caekenberghe <[hidden email]> wrote: > > > On 26 Jan 2016, at 16:49, Dimitris Chloupis <[hidden email]> wrote: > > > > Obviously it will better fit Pharo since its made to work with smalltalk code, but that does not make it any less terrible. Just because you have one implementation of something that does not mean its good. Its just means its there and it works. > > > > I dont know the internal, they are not documented anyway, there are some class comments here and there but thats pretty much it. I dont even remember when was the last time monticello got an updated, I mean a serious update not just a couple of bug fixes the 2 years I have been around. > > > > Secondly GUI is just plain awful, Smalltalk maybe be the first or one of the first to implement guis, but those implementations never ended up to something that would be approachable and easy to use on a day to day basis, some tools suffer more from this some less, Monticello is up there with the worse design. > > > > Thirdly the inability of the system to version control images , audio files and other assets it defeats the central purpose of smalltalk of everything being objects with a loud "Nope !" from Monticello "Only source code is". > > > > So its awesome that Smalltalk , and Squeak got its own version control system, that is easy to use and Pharo inherited it. Congratulations to people behind it. But the GUI needs to go, its a bad advertisement to Pharo, and we need something that is not stuck to dark ages as you correctly pointed but for the opposite reason. Because any way you try to turn Monticello you wont find a label written "modern" on it. The label you may find on it is more like "abandonware". > > > > Also there like a ton of OOP languages out there using git with no major problems, the problem with smalltalk is that smalltalk is an island. > > > > And the problem with islands is when you end having fun with them you feel stuck since they dont provide an easy access to the outside world. > > > > "Git just manages blobs, text files at best. Dead text" > > > > Last time I checked Monticello used a format called mcz which is nothing more than a zip file containing st files, which are as you call it "dead text" files. Also I would like to remind you that git is used by the CUIS smalltalk to version control their images, I thought images are live code. > > > > Personally I dont see the diffirence between live and dead text. Its just text to me. The VM is the one that makes it live anyway. > > Yes and no. > > If git compares two versions, it does not understand what is in it. > > When Monticello compares versions, it knows about classes, methods, inheritance, > it can explain diffs in the same structured (browsable) form that you wrote them. > > Putting Smalltalk in plain text files can be done, has been done, and was not successful, because you lose something essential by leaving the environment. > > The island argument is also a bit too easy: any platform can be seen as an island that needs to integrate with others, there are degrees of openness. Your Blender or Python are islands too, just like SAP, Microsoft and Oracle. Some connections/interfaces are easy, some are hard(er). It all depends on your position. > > When have you last looked at the source code of git ? > > And be honest, do you find git always easy to use ? > > > On Tue, Jan 26, 2016 at 5:09 PM Sven Van Caekenberghe <[hidden email]> wrote: > > > > > On 26 Jan 2016, at 15:59, Dimitris Chloupis <[hidden email]> wrote: > > > > > > To be fair my experience with pharo and git have been not always smooth either. I have the VM crashing again and again completely randomly when it was trying to pull SmaCC as dependency for my project Ephestos, had to drop SmaCC and moving my python type parsing at python side. > > > > > > But I find it ironic someone using Monticello, trying to equate git with dark ages, you cant get more dark ages than monticello, frankly. No offense to people who made it , its great that is in there but its full of problems and bad designs and cant even begin to be compared with Github and GIT GUI clients. Monticello is according to my personal opinion by far the worst tool of Pharo. > > > > No it is not. It is a version management system that understands our object and code model, a system that we control. Git just manages blobs, text files at best. Dead text. > > > > (This does not mean it is perfect, nor that it cannot improve, nor that we should not improve our git integration.) > > > > > On Tue, Jan 26, 2016 at 4:51 PM Thierry Goubier <[hidden email]> wrote: > > > 2016-01-26 15:11 GMT+01:00 Sean P. DeNigris <[hidden email]>: > > > NorbertHartl wrote > > > > - I need to use BaselineOf instead of ConfigurationOf. Thus you cannot use > > > > Versionner anymore > > > > > > Unfortunately. This is the biggest drag for me after switching all my > > > personal projects to git (GitHub for public and BitBucket for private). I > > > had gotten spoiled by Versionner and hand-editing MetaC artifacts feels like > > > going back to the dark ages :/ > > > > > > Well, I guess copying the baselines generated by Versionner into a BaselineOf is probably a way to do it. > > > > > > Thierry > > > > > > > > > > > > > > > ----- > > > Cheers, > > > Sean > > > -- > > > View this message in context: http://forum.world.st/Pharo-git-workflow-tp4874067p4874221.html > > > Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com. > > > > > > > > > |
|
|
I love live coding , dynamic language, beautiful syntax and close integration with IDE. But yeah Monticello , I don't get it. I love Pharo and always respect the hard work of others but that does not mean I love every part of it. Same story is Blender , Python and all other tools I use. That ok I don't worry if some parts I don't like. It's what make with the tools you have that matters most.
On Tue, 26 Jan 2016 at 19:46, Sven Van Caekenberghe <[hidden email]> wrote: Sometimes I really don't understand what you see in Pharo ... |
|
|
In reply to this post by NorbertHartl
Hi Norbert,
the development version on Pharo4 has the credentials working now. It switches to https instead of ssh if you add a username and a password, since a ssh url can't carry a password and, in the case of github, the ssh user is allways git. Thierry Le 26/01/2016 13:32, Norbert Hartl a écrit : > >> Am 26.01.2016 um 13:03 schrieb Norbert Hartl <[hidden email] >> <mailto:[hidden email]>>: >> >>> >>> Am 25.01.2016 um 23:32 schrieb Thierry Goubier >>> <[hidden email] <mailto:[hidden email]>>: >>> >>> Le 25/01/2016 23:13, Norbert Hartl a écrit : >>>> >>>>> Am 25.01.2016 um 23:02 schrieb Norbert Hartl <[hidden email] >>>>> <mailto:[hidden email]>>: >>>>> >>>>> Hi, >>>>> >>>>>> Am 25.01.2016 um 22:55 schrieb Thierry Goubier >>>>>> <[hidden email] <mailto:[hidden email]>>: >>>>>> >>>>>> Hi Norbert, >>>>>> >>>>>> Just tell me if you need any additional parameter to the git clone >>>>>> for the credentials, because I'm not sure I have written the code >>>>>> which will handle them. >>>>>> >>>>> I'll check that. The easiest approach is to use the authority part >>>>> of the url like >>>>> >>>>> https://user:pass@.../…. >>> >>> I'm nearly sure my code doesn't check for them and silently drop them :( >>> >>>>> I try that tomorrow. >>>>> >>>> But then I don't think that sub-directories work per http. >>> >>> They do over https. But remember the syntax is, to take your url, >>> >>> gitfiletree://user:[hidden email]/...?protocol=https >> >> Does not work for me. It complains about unknown url scheme. But using >> the read-only version of gitfiletree would still mean I have to >> install gitfiletree, no? I have no glue hot the url with bitbucket:// >> could work. >>> >>> I wonder if I can write code that considers that if you have a >>> username and a password, then it should use https and not ssh. >> >> I see no reason why. You can use ssh with username and password as well. >> > Ok, I got it through the unknown url scheme but indeed the credentials > are stripped off. > > Norbert > |
|
|
Hi Thierry,
> Am 27.01.2016 um 06:40 schrieb Thierry Goubier <[hidden email]>: > > Hi Norbert, > > the development version on Pharo4 has the credentials working now. It switches to https instead of ssh if you add a username and a password, since a ssh url can't carry a password and, in the case of github, the ssh user is allways git. > thanks very much. I need to find some time to try it out. I have kind of a workflow but that does only work if I use a single repository. Norbert > Thierry > > Le 26/01/2016 13:32, Norbert Hartl a écrit : >> >>> Am 26.01.2016 um 13:03 schrieb Norbert Hartl <[hidden email] >>> <mailto:[hidden email]>>: >>> >>>> >>>> Am 25.01.2016 um 23:32 schrieb Thierry Goubier >>>> <[hidden email] <mailto:[hidden email]>>: >>>> >>>> Le 25/01/2016 23:13, Norbert Hartl a écrit : >>>>> >>>>>> Am 25.01.2016 um 23:02 schrieb Norbert Hartl <[hidden email] >>>>>> <mailto:[hidden email]>>: >>>>>> >>>>>> Hi, >>>>>> >>>>>>> Am 25.01.2016 um 22:55 schrieb Thierry Goubier >>>>>>> <[hidden email] <mailto:[hidden email]>>: >>>>>>> >>>>>>> Hi Norbert, >>>>>>> >>>>>>> Just tell me if you need any additional parameter to the git clone >>>>>>> for the credentials, because I'm not sure I have written the code >>>>>>> which will handle them. >>>>>>> >>>>>> I'll check that. The easiest approach is to use the authority part >>>>>> of the url like >>>>>> >>>>>> https://user:pass@.../…. >>>> >>>> I'm nearly sure my code doesn't check for them and silently drop them :( >>>> >>>>>> I try that tomorrow. >>>>>> >>>>> But then I don't think that sub-directories work per http. >>>> >>>> They do over https. But remember the syntax is, to take your url, >>>> >>>> gitfiletree://user:[hidden email]/...?protocol=https >>> >>> Does not work for me. It complains about unknown url scheme. But using >>> the read-only version of gitfiletree would still mean I have to >>> install gitfiletree, no? I have no glue hot the url with bitbucket:// >>> could work. >>>> >>>> I wonder if I can write code that considers that if you have a >>>> username and a password, then it should use https and not ssh. >>> >>> I see no reason why. You can use ssh with username and password as well. >>> >> Ok, I got it through the unknown url scheme but indeed the credentials >> are stripped off. >> >> Norbert >> > > |
Re: Pharo + git workflow
|
|
In reply to this post by kilon.alios
For me what monticello has and git not is that is unobtrusive. Yes
interface is not pretty, but once you have the basics it becomes
invisible. Not the case with git. Hopefully the DVCS support will
have the same amount of invisibility. I know that now things are
moving, but long threads about the integration of git + pharo show
the responsiveness of the community in one hand and the lot of
support it needs for working on the other.
Cheers, Offray On 26/01/16 13:10, Dimitris Chloupis
wrote:
I love live coding , dynamic language, beautiful syntax and close integration with IDE. But yeah Monticello , I don't get it. I love Pharo and always respect the hard work of others but that does not mean I love every part of it. Same story is Blender , Python and all other tools I use. That ok I don't worry if some parts I don't like. It's what make with the tools you have that matters most. |
|
|
In reply to this post by NorbertHartl
Finally I got something working for me. It is not generally applicable but at the moment I'm happy this works at least. I made it even harder because I store the smalltalk code in a sub directory because I want to version web resources along with the code. The smalltalk code is in a sub directory st/
development desktop - - - - - - - - - - - - - - - - Normal GitFileTree workflow: - Take a pharo4 image and load the configuration of GitFileTree from the Configuration Browser. - Open Versionner then select GitFileTree and click on development and then load version (via context menu). This is IMHO only needed if you want to access bitbucket - Checkout your repository - Create GitFileTree repository and select directory of checked out repository - If a ssh key is configured in bitbucket and for the local ssh-agent commit, pull, push should work via monticello browser jenkins - - - - - In the jenkins build I like to avoid GitFileTree because I prefer having neither GitFileTree nor OSProcess installed in my deployment artefacts - Configure jenkins git plugin to check out the code repository - add a shell command consisting pharo-vm-nox project.image config "filetree://$WORKSPACE/st" ConfigurationOfProject --install=bleedingEdge - this has access to the ConfigurationOfProject but only to this class - Add spec repository: 'filetree://st/' to the baseline of the ConfigurationOfProject. This way metacello can load the rest of the packages. The good news is that I can have a full powered GitFileTree enabled version for developing and a OSProcess-less version for deployment. And I can use Versionner as I rely on the ConfigurationOf. But especially the spec repository setting in the ConfigurationOfProject is a bad hack. Another solution to this would be welcomed. This works only if all packages are in the same git repository. I need to try access that stuff via bitbucket:// url in the metacello baseline. Thierry just added support to put credentials in the url so it might work with downloading read-only versions of the repository. Norbert > Am 25.01.2016 um 18:09 schrieb Norbert Hartl <[hidden email]>: > > I'm eager to try a new project with some git repositories. But to be honest I don't really get it. Searching the web there is lots to find but nothing actual. > > I don't understand if it is ok to use a ConfigurationOf or if it only works with a BaselineOf. And how do you specify the repository in a ConfigurationOf in order to be able to work locally as well as having jenkins pull everything automatically? Same goes for dependent projects. > > Are there any insights to this or pointers to an up-to-date documentation? > > thanks, > > Norbert |
|
|
Norbert,
If you were using a BaselineOf for your project then you could avoid the the configuration hack by using the Metacello lock command[1] which was designed to address the problem where you want to use a local git clone while specifying project references in the configurations using the github (or bitbucket) repository url.... I understand that BaselineOf support is missing from versionner, but until the rest of the tool chain is in place, you will have to trade off where you want to "insert hacks" ... it may be easier to hand edit the BaselineOf when adding a new package or project dependency, than constantly hack configurations because you aren't using a BaselineOf --- but then that's up to you:) Dale [1] https://github.com/dalehenrich/metacello-work/blob/master/docs/LockCommandReference.md On 01/28/2016 09:39 AM, Norbert Hartl wrote: > Finally I got something working for me. It is not generally applicable but at the moment I'm happy this works at least. I made it even harder because I store the smalltalk code in a sub directory because I want to version web resources along with the code. The smalltalk code is in a sub directory st/ > > development desktop > - - - - - - - - - - - - - - - - > > Normal GitFileTree workflow: > > - Take a pharo4 image and load the configuration of GitFileTree from the Configuration Browser. > - Open Versionner then select GitFileTree and click on development and then load version (via context menu). This is IMHO only needed if you want to access bitbucket > - Checkout your repository > - Create GitFileTree repository and select directory of checked out repository > - If a ssh key is configured in bitbucket and for the local ssh-agent commit, pull, push should work via monticello browser > > jenkins > - - - - - > > In the jenkins build I like to avoid GitFileTree because I prefer having neither GitFileTree nor OSProcess installed in my deployment artefacts > > - Configure jenkins git plugin to check out the code repository > - add a shell command consisting > > pharo-vm-nox project.image config "filetree://$WORKSPACE/st" ConfigurationOfProject --install=bleedingEdge > > - this has access to the ConfigurationOfProject but only to this class > - Add > > spec repository: 'filetree://st/' > > to the baseline of the ConfigurationOfProject. This way metacello can load the rest of the packages. > > The good news is that I can have a full powered GitFileTree enabled version for developing and a OSProcess-less version for deployment. And I can use Versionner as I rely on the ConfigurationOf. But especially the spec repository setting in the ConfigurationOfProject is a bad hack. Another solution to this would be welcomed. This works only if all packages are in the same git repository. > I need to try access that stuff via bitbucket:// url in the metacello baseline. Thierry just added support to put credentials in the url so it might work with downloading read-only versions of the repository. > > Norbert > >> Am 25.01.2016 um 18:09 schrieb Norbert Hartl <[hidden email]>: >> >> I'm eager to try a new project with some git repositories. But to be honest I don't really get it. Searching the web there is lots to find but nothing actual. >> >> I don't understand if it is ok to use a ConfigurationOf or if it only works with a BaselineOf. And how do you specify the repository in a ConfigurationOf in order to be able to work locally as well as having jenkins pull everything automatically? Same goes for dependent projects. >> >> Are there any insights to this or pointers to an up-to-date documentation? >> >> thanks, >> >> Norbert > |
«
Return to Pharo Smalltalk Users
|
1 view|%1 views
| Free forum by Nabble | Edit this page |