On Fri, Apr 15, 2011 at 11:49:30PM +0200, Levente Uzonyi wrote:
> On Fri, 15 Apr 2011, David T. Lewis wrote: > > >Thanks Casey, > > > >I'm not sure how to make this better (and my point was really only > >to focus more on how to make it better versus how to nag people etc). > > > >That said, I think that the problem I experience is that I have a > >hard time looking at something in the inbox, figuring out how out > >of date it might be, and figuring out exactly what changes it was > >originally attempting to make. In many cases, the submission might > >involve just a few methods, but it takes me a long time to figure > >that out by browsing the MCZ and cross-checking against emails. > > > >I find the Montecello process to be wonderful for development and > >for maintaining the update stream, but when I look at something > >that someone else submitted a few weeks ago, I find myself wishing > >that I could just look at the change set. > > > >So maybe I am just looking for a button that says "show me the > >change set" where the change set would be the changes that the > >original author was submitting two weeks ago. > > For a single package, use MC's merge button. It will show all changes > and even highlight conflicts with newer changes in the Trunk. > > For multiple packages, we have no solution yet. What you can do is to > merge all packages into your image and then create a changeset from the > changesets created by MC, using the Dual Change Sorter. > > > > >I have an uneasy feeling that there is some existing way to do > >this and I'm too dumb to have noticed it yet, so I'm preparing > >myself for an embarassing reply from Bert within the next few > >minutes ;) > > Bert modified our Squeaksource to create diffs (like those which are sent > to this list) and changesets, but changesets only work for packages > already in the Trunk for some reason. Levente, Thanks for the tips. And of course Bert's SqueakSource code is all on source.squeak.org, but I never thought to look there (d'oh!). Dave |
In reply to this post by Casey Ransberger-2
On Fri, Apr 15, 2011 at 3:09 PM, Casey Ransberger <[hidden email]> wrote: I think most existing SCM solutions make this less painful with magic branching/merging/digging goodness. MC doesn't support branches, so we use a separate repository, which I think kind of sucks. I doubt we can fix that easily, though. Well, with VMMaker we're using an arguably dubious convention that supports branching Cog from the Interpreter in the same repository just fine. So I'm not sure I agree. 2¢ Eliot
|
You mean the MC package, not the SVN sources right? What's the convention?
On Fri, Apr 15, 2011 at 9:21 PM, Eliot Miranda <[hidden email]> wrote:
-- Casey Ransberger |
On Fri, Apr 15, 2011 at 10:49 PM, Casey Ransberger <[hidden email]> wrote: You mean the MC package, not the SVN sources right? Right. What's the convention? Use some false initials that name the branch, and start its version numbers from e.g. 1. So I name my open-source Cog versions VMMaker-oscog.N. The package still has my initials as author. David is using VMMaker-oscog.dtl.N, etc.
|
On Fri, Apr 15, 2011 at 10:57 PM, Eliot Miranda <[hidden email]> wrote:
> Use some false initials that name the branch, and start its version numbers > from e.g. 1. So I name my open-source Cog versions VMMaker-oscog.N. The > package still has my initials as author. David is using > VMMaker-oscog.dtl.N, etc. Actually, MC does have some support for branching. If you use the following convention, MC will do the "right" thing - group the branched versions together in the UI, automatically suggest a version name that's on a branch, and still allow you to merge between branches. The naming convention is: <package>.<branch>-<initials>.<id> So if Eliot were using the standard convention, his versions would be named like this: VMMaker.oscog-eem.147 Colin |
On Sat, Apr 16, 2011 at 9:37 AM, Colin Putney <[hidden email]> wrote:
I shall start using this. Thanks. Eliot
|
On 04/16/2011 06:42 PM, Eliot Miranda wrote:
> > > On Sat, Apr 16, 2011 at 9:37 AM, Colin Putney <[hidden email] > <mailto:[hidden email]>> wrote: > > On Fri, Apr 15, 2011 at 10:57 PM, Eliot Miranda > <[hidden email] <mailto:[hidden email]>> wrote: > > > Use some false initials that name the branch, and start its > version numbers > > from e.g. 1. So I name my open-source Cog versions > VMMaker-oscog.N. The > > package still has my initials as author. David is using > > VMMaker-oscog.dtl.N, etc. > > Actually, MC does have some support for branching. If you use the > following convention, MC will do the "right" thing - group the > branched versions together in the UI, automatically suggest a version > name that's on a branch, and still allow you to merge between > branches. The naming convention is: > > <package>.<branch>-<initials>.<id> > > So if Eliot were using the standard convention, his versions would be > named like this: > > VMMaker.oscog-eem.147 > > > I shall start using this. Thanks. We used a similar scheme in Gjallar development - works great. And one branch/person can be designated as "trunk" and merges the other's snapshots in selectively and typically makes snapshots that "work". Then the other devs can just snapshot away (possibly by first rebasing, but we didn't do that) and occasionally merge with the "trunk" to get all the other's contributions that have been allowed into trunk. regards, Göran |
In reply to this post by Colin Putney-3
Interesting naming scheme; is that something we should "document" in
MCVersionName? Also, just to clarify our words, MC _does_ support branches in the domain, just not names for them. But in terms of the ancestry, you can unlimited branching, which could be rendered if the tools were improved to do that.. Adding a name to something can often be done via external annotation, notwithstanding this naming scheme.. On Sat, Apr 16, 2011 at 11:37 AM, Colin Putney <[hidden email]> wrote: > On Fri, Apr 15, 2011 at 10:57 PM, Eliot Miranda <[hidden email]> wrote: > >> Use some false initials that name the branch, and start its version numbers >> from e.g. 1. So I name my open-source Cog versions VMMaker-oscog.N. The >> package still has my initials as author. David is using >> VMMaker-oscog.dtl.N, etc. > > Actually, MC does have some support for branching. If you use the > following convention, MC will do the "right" thing - group the > branched versions together in the UI, automatically suggest a version > name that's on a branch, and still allow you to merge between > branches. The naming convention is: > > <package>.<branch>-<initials>.<id> > > So if Eliot were using the standard convention, his versions would be > named like this: > > VMMaker.oscog-eem.147 > > Colin > > |
Well, I've learned something today! I wonder if I can fit my other foot in my mouth at the same time...
On Sat, Apr 16, 2011 at 12:33 PM, Chris Muller <[hidden email]> wrote: Interesting naming scheme; is that something we should "document" in -- Casey Ransberger |
In reply to this post by Chris Muller-3
On Sat, Apr 16, 2011 at 12:33 PM, Chris Muller <[hidden email]> wrote:
> Interesting naming scheme; is that something we should "document" in > MCVersionName? > > Also, just to clarify our words, MC _does_ support branches in the > domain, just not names for them. But in terms of the ancestry, you > can unlimited branching, which could be rendered if the tools were > improved to do that.. Adding a name to something can often be done > via external annotation, notwithstanding this naming scheme.. Well, actually, it supports names for branches as well. The problem is that it doesn't have a "branch" command, so there's no way to discover the naming convention for branches. You just have to know. |
Is there anything blocking the addition of some UI for it? Or was that just a YAGNI thing?
It'd be neat to expose the feature to folks like me who don't know the code base:) On Apr 16, 2011, at 2:23 PM, Colin Putney <[hidden email]> wrote: > On Sat, Apr 16, 2011 at 12:33 PM, Chris Muller <[hidden email]> wrote: >> Interesting naming scheme; is that something we should "document" in >> MCVersionName? >> >> Also, just to clarify our words, MC _does_ support branches in the >> domain, just not names for them. But in terms of the ancestry, you >> can unlimited branching, which could be rendered if the tools were >> improved to do that.. Adding a name to something can often be done >> via external annotation, notwithstanding this naming scheme.. > > Well, actually, it supports names for branches as well. The problem is > that it doesn't have a "branch" command, so there's no way to discover > the naming convention for branches. You just have to know. > |
On Sat, Apr 16, 2011 at 2:32 PM, Casey Ransberger
<[hidden email]> wrote: > Is there anything blocking the addition of some UI for it? Or was that just a YAGNI thing? I think it was mostly that we *didn't* need it because MC lets you edit the name of a version anyway. Shouldn't be a problem at all. Colin |
In reply to this post by Chris Muller-3
On 4/16/11, Chris Muller <[hidden email]> wrote:
> Interesting naming scheme; is that something we should "document" in > MCVersionName? Yes, please --Hannes > > Also, just to clarify our words, MC _does_ support branches in the > domain, just not names for them. But in terms of the ancestry, you > can unlimited branching, which could be rendered if the tools were > improved to do that.. Adding a name to something can often be done > via external annotation, notwithstanding this naming scheme.. > > > On Sat, Apr 16, 2011 at 11:37 AM, Colin Putney <[hidden email]> wrote: >> On Fri, Apr 15, 2011 at 10:57 PM, Eliot Miranda <[hidden email]> >> wrote: >> >>> Use some false initials that name the branch, and start its version >>> numbers >>> from e.g. 1. So I name my open-source Cog versions VMMaker-oscog.N. The >>> package still has my initials as author. David is using >>> VMMaker-oscog.dtl.N, etc. >> >> Actually, MC does have some support for branching. If you use the >> following convention, MC will do the "right" thing - group the >> branched versions together in the UI, automatically suggest a version >> name that's on a branch, and still allow you to merge between >> branches. The naming convention is: >> >> <package>.<branch>-<initials>.<id> >> >> So if Eliot were using the standard convention, his versions would be >> named like this: >> >> VMMaker.oscog-eem.147 >> >> Colin >> >> > > |
On Sun, 17 Apr 2011, Hannes Hirzel wrote:
> On 4/16/11, Chris Muller <[hidden email]> wrote: >> Interesting naming scheme; is that something we should "document" in >> MCVersionName? > > Yes, please The question was not about if there should be some documentation added or not, but whether if this naming convetion is/should be widely used. There's at least one other existing naming scheme for branches (http://wiki.squeak.org/squeak/3328 ), but IIRC we used to use something for Gjallar too. So IMHO if it's added to the class comment, then it should only be a suggestion for future users. Levente > > --Hannes > >> >> Also, just to clarify our words, MC _does_ support branches in the >> domain, just not names for them. But in terms of the ancestry, you >> can unlimited branching, which could be rendered if the tools were >> improved to do that.. Adding a name to something can often be done >> via external annotation, notwithstanding this naming scheme.. >> >> >> On Sat, Apr 16, 2011 at 11:37 AM, Colin Putney <[hidden email]> wrote: >>> On Fri, Apr 15, 2011 at 10:57 PM, Eliot Miranda <[hidden email]> >>> wrote: >>> >>>> Use some false initials that name the branch, and start its version >>>> numbers >>>> from e.g. 1. So I name my open-source Cog versions VMMaker-oscog.N. The >>>> package still has my initials as author. David is using >>>> VMMaker-oscog.dtl.N, etc. >>> >>> Actually, MC does have some support for branching. If you use the >>> following convention, MC will do the "right" thing - group the >>> branched versions together in the UI, automatically suggest a version >>> name that's on a branch, and still allow you to merge between >>> branches. The naming convention is: >>> >>> <package>.<branch>-<initials>.<id> >>> >>> So if Eliot were using the standard convention, his versions would be >>> named like this: >>> >>> VMMaker.oscog-eem.147 >>> >>> Colin >>> >>> >> >> > > |
Personally, I think we should leave the names alone and let the
'ancestry' determine the... ancestry. Because otherwise the names could get "out of sync" with real ancestry; or some tools of the future would invariably rely on parsing names (yuck) to determine ancestry rather than the actual domain. I, personally, don't think such naming is that great of a benefit; because whereever there are majorly-different branches of a package, they are worked on in different *images*; and they don't even have to be in separate MC repositories at all, even with FileBased; because you can see the real History straight from the domain object (MCVersionInfo) via the "History" button, which excludes all of the versions from the "other" branch. So I don't see that complicated naming solves any real problem.. - Chris On Sun, Apr 17, 2011 at 11:44 AM, Levente Uzonyi <[hidden email]> wrote: > On Sun, 17 Apr 2011, Hannes Hirzel wrote: > >> On 4/16/11, Chris Muller <[hidden email]> wrote: >>> >>> Interesting naming scheme; is that something we should "document" in >>> MCVersionName? >> >> Yes, please > > The question was not about if there should be some documentation added or > not, but whether if this naming convetion is/should be widely used. There's > at least one other existing naming scheme for branches > (http://wiki.squeak.org/squeak/3328 ), but IIRC we used to use something for > Gjallar too. > So IMHO if it's added to the class comment, then it should only be a > suggestion for future users. > > > Levente > >> >> --Hannes >> >>> >>> Also, just to clarify our words, MC _does_ support branches in the >>> domain, just not names for them. But in terms of the ancestry, you >>> can unlimited branching, which could be rendered if the tools were >>> improved to do that.. Adding a name to something can often be done >>> via external annotation, notwithstanding this naming scheme.. >>> >>> >>> On Sat, Apr 16, 2011 at 11:37 AM, Colin Putney <[hidden email]> >>> wrote: >>>> >>>> On Fri, Apr 15, 2011 at 10:57 PM, Eliot Miranda >>>> <[hidden email]> >>>> wrote: >>>> >>>>> Use some false initials that name the branch, and start its version >>>>> numbers >>>>> from e.g. 1. So I name my open-source Cog versions VMMaker-oscog.N. >>>>> The >>>>> package still has my initials as author. David is using >>>>> VMMaker-oscog.dtl.N, etc. >>>> >>>> Actually, MC does have some support for branching. If you use the >>>> following convention, MC will do the "right" thing - group the >>>> branched versions together in the UI, automatically suggest a version >>>> name that's on a branch, and still allow you to merge between >>>> branches. The naming convention is: >>>> >>>> <package>.<branch>-<initials>.<id> >>>> >>>> So if Eliot were using the standard convention, his versions would be >>>> named like this: >>>> >>>> VMMaker.oscog-eem.147 >>>> >>>> Colin >>>> >>>> >>> >>> >> >> > > |
On Sun, Apr 17, 2011 at 10:19 AM, Chris Muller <[hidden email]> wrote:
> Personally, I think we should leave the names alone and let the > 'ancestry' determine the... ancestry. Names don't determine the ancestry. The only thing the name is used for is display in the UI. And that's all a "branch" is in MC - a bunch of versions that are grouped together in the UI. Colin |
In reply to this post by Chris Muller-3
On Sun, Apr 17, 2011 at 12:19:49PM -0500, Chris Muller wrote:
> > On Sun, Apr 17, 2011 at 11:44 AM, Levente Uzonyi <[hidden email]> wrote: > > On Sun, 17 Apr 2011, Hannes Hirzel wrote: > > > >> On 4/16/11, Chris Muller <[hidden email]> wrote: > >>> > >>> Interesting naming scheme; is that something we should "document" in > >>> MCVersionName? > >> > >> Yes, please > > > > The question was not about if there should be some documentation added or > > not, but whether if this naming convetion is/should be widely used. There's > > at least one other existing naming scheme for branches > > (http://wiki.squeak.org/squeak/3328 ), but IIRC we used to use something for > > Gjallar too. > > So IMHO if it's added to the class comment, then it should only be a > > suggestion for future users. > > > Personally, I think we should leave the names alone and let the > 'ancestry' determine the... ancestry. > > Because otherwise the names could get "out of sync" with real > ancestry; or some tools of the future would invariably rely on parsing > names (yuck) to determine ancestry rather than the actual domain. > > I, personally, don't think such naming is that great of a benefit; > because whereever there are majorly-different branches of a package, > they are worked on in different *images*; and they don't even have to > be in separate MC repositories at all, even with FileBased; because > you can see the real History straight from the domain object > (MCVersionInfo) via the "History" button, which excludes all of the > versions from the "other" branch. So I don't see that complicated > naming solves any real problem.. Actually it does solve a real problem. Between Eliot, Igor, and myself, not one of us knew enough to use a good naming convention in the VMMaker project. Thus we missed an opportunity to have the branches displayed more clearly in the browser. So even if this is just a naming hack, and even if it should only be taken as a suggestion for future users, it still would be helpful to document it. I agree with Hannes: "yes please!" Dave |
> Actually it does solve a real problem. Between Eliot, Igor, and myself,
> not one of us knew enough to use a good naming convention in the VMMaker > project. Thus we missed an opportunity to have the branches displayed > more clearly in the browser. Ok, but that reinforces my point that it's a tool shortfall that can't display the _real_ branches; you had to rely on filename sorting to see the individual branches clearly. But if one of you had accidently used the naming convention of one of your peers on your own branch, MC would "incorrectly" display that version intermingled in your own "branch", because the real branch is determined by the ancestry, not the name. Colin wrote: > And that's all a "branch" is in MC - a bunch > of versions that are grouped together in the UI. No, a branch is based on the ancestry, regardless of how they're grouped in the UI. I understand we WANT separate branches to be grouped in the UI, but currently they aren't. I think it would be nice if the tools could do it; for example, maybe a hierarchy where only the most-current "leaf" versions of each branch are displayed, but they could be expanded to show their list of immediate ancestors. > So even if this is just a naming hack, > and even if it should only be taken as a suggestion for future users, > it still would be helpful to document it. I think it should be "documented" as accessing methods in MCVersionName. - Chris |
> would "incorrectly" display that version intermingled in your own
I meant to say, "intermingled in the other guys list", not your own. |
In reply to this post by Chris Muller-4
Chris Muller wrote (among other things):
> it's a tool shortfall that can't display the _real_ branches and: > No, a branch is based on the ancestry, regardless of how they're > grouped in the UI. You seem to have some pretty strong opinions about branches. Would you care to expand on this a little bit? Maybe we can learn something. Colin |
Free forum by Nabble | Edit this page |