{"thread":{"id":"21741","subject":"git-subtree: directory mismatch","startedAt":"2009-11-24T19:53:53Z","lastAt":"2009-11-25T23:45:22Z","messageCount":12,"participants":["Marc Fournier","Avery Pennarun","Nanako Shiraishi","Junio C Hamano","Johannes Schindelin","Björn Steinbrink"],"isPatch":false,"patchVersion":null,"patchTotal":null},"messages":[{"id":"128260","messageId":"20091124195353.GA16627@lonquimay.wrk.lsn.camptocamp.com","threadId":"21741","inReplyTo":null,"subject":"git-subtree: directory mismatch","fromName":"Marc Fournier","fromEmail":"marc.fournier@camptocamp.com","sentAt":"2009-11-24T19:53:53Z","receivedAt":"2009-11-24T19:53:53Z","isPatch":false,"sender":{"key":"marc.fournier@camptocamp.com","avatar":null},"body":"Hi,\n\nI've come across this weird behaviour using git-subtree: when updating a\ndirectory created by git-subtree, the content of some (almost) random\nneighbour directory gets replaced, instead of the directory specified by\n\"git-subtree --prefix\".\n\n\nmkdir /tmp/testrepo\ncd /tmp/testrepo\ngit init .\nmkdir -p modules/cognac/{manifests,files}\nmkdir -p modules/cognac/manifests/{classes,definitions}\ntouch modules/cognac/README modules/cognac/manifests/classes/foo.pp\nmodules/cognac/manifests/definitions/bar.pp modules/cognac/files/blabla\ngit add modules/\ngit commit -m \"test\"\n\ngit remote add mapserver git://github.com/camptocamp/puppet-mapserver.git\ngit remote update\ngit-subtree add --prefix modules/mapserver 97f5985b # an older commit\n\n# So far so good, I have 2 subdirectories in modules/ with completely\n# different content.\n\n# Then, the content of modules/cognac/ gets completely replaced by the\n# content of HEAD from git://github.com/camptocamp/puppet-mapserver.git.\n# modules/mapserver still contains the content of commit 97f5985b.\n\ngit-subtree pull --prefix modules/mapserver mapserver master\n\n>From git://github.com/camptocamp/puppet-mapserver\n * branch master -> FETCH_HEAD\nRemoving modules/cognac/files/blabla\nRemoving modules/cognac/manifests/classes/foo.pp\nRemoving modules/cognac/manifests/definitions/bar.pp\nMerge made by subtree.\n modules/cognac/files/apt-preferences | 3 +\n modules/cognac/files/epsg.C2C | 5445 ++++++++++++++++++++\n modules/cognac/files/epsg.legacy | 5445 ++++++++++++++++++++\n modules/cognac/files/epsg.minimal | 14 +\n modules/cognac/files/etc/apt/preferences-20070910 | 3 +\n modules/cognac/files/etc/apt/preferences-20080225 | 35 +\n modules/cognac/files/etc/apt/preferences-v5-2 | 35 +\n modules/cognac/manifests/classes/epsg.pp | 21 +\n .../cognac/manifests/classes/mapserver-debian.pp | 82 +\n modules/cognac/manifests/classes/v5.pp | 46 +\n modules/cognac/manifests/init.pp | 8 +\n 11 files changed, 11137 insertions(+), 0 deletions(-)\n create mode 100644 modules/cognac/files/apt-preferences\n delete mode 100644 modules/cognac/files/blabla\n create mode 100755 modules/cognac/files/epsg.C2C\n create mode 100755 modules/cognac/files/epsg.legacy\n create mode 100644 modules/cognac/files/epsg.minimal\n create mode 100644 modules/cognac/files/etc/apt/preferences-20070910\n create mode 100644 modules/cognac/files/etc/apt/preferences-20080225\n create mode 100644 modules/cognac/files/etc/apt/preferences-v5-2\n create mode 100644 modules/cognac/manifests/classes/epsg.pp\n delete mode 100644 modules/cognac/manifests/classes/foo.pp\n create mode 100644 modules/cognac/manifests/classes/mapserver-debian.pp\n create mode 100644 modules/cognac/manifests/classes/v5.pp\n delete mode 100644 modules/cognac/manifests/definitions/bar.pp\n create mode 100644 modules/cognac/manifests/init.pp\n\nI was not able to reproduce this bug in any of the following cases:\n - if the README file wasn't empty\n - if modules/cognac is renamed to something else\n - if modules/cognac doesn't have the same sub-directories than\n modules/mapserver\n\nThis bug happens using git 1.5.5.6 as well as 1.6.5. It seems to happen\nwhen git-subtree calls \"git merge -s subtree\".\n\nMaybe something is broken in this puppet-mapserver.git repository I use ?\nI've encountered the same problem with other directories too.\n\nAny idea ?\n\nThanks !\nMarc\n"},{"id":"128263","messageId":"32541b130911241348s21e21fb8n12edf374e6a3c309@mail.gmail.com","threadId":"21741","inReplyTo":"20091124195353.GA16627@lonquimay.wrk.lsn.camptocamp.com","subject":"Re: git-subtree: directory mismatch","fromName":"Avery Pennarun","fromEmail":"apenwarr@gmail.com","sentAt":"2009-11-24T21:48:26Z","receivedAt":"2009-11-24T21:48:26Z","isPatch":false,"sender":{"key":"apenwarr@gmail.com","avatar":"https://avatars.githubusercontent.com/u/20592?v=4"},"body":"On Tue, Nov 24, 2009 at 2:53 PM, Marc Fournier\n wrote:\n> I was not able to reproduce this bug in any of the following cases:\n>  - if the README file wasn't empty\n>  - if modules/cognac is renamed to something else\n>  - if modules/cognac doesn't have the same sub-directories than\n>   modules/mapserver\n>\n> This bug happens using git 1.5.5.6 as well as 1.6.5. It seems to happen\n> when git-subtree calls \"git merge -s subtree\".\n\nYup. This is basically a bug in \"git merge -s subtree\": it guesses\nwhich subtree to merge into, rather than actually taking a prefix\nparameter. I've been meaning to either submit a patch for this, or\nfind a way to work around it.\n\nThis doesn't usually happen once your project is relatively mature\n(ie. doesn't have blank or \"default\" template files in it) since then\nthe auto-guessing gets more reliable. But there's no good reason to\ndo the auto-guessing, so it would be best to do this \"properly.\"\n\nSorry that I haven't had time to fix it yet...\n\nAvery\n"},{"id":"128267","messageId":"20091125080812.6117@nanako3.lavabit.com","threadId":"21741","inReplyTo":"32541b130911241348s21e21fb8n12edf374e6a3c309@mail.gmail.com","subject":"Re: git-subtree: directory mismatch","fromName":"Nanako Shiraishi","fromEmail":"nanako3@lavabit.com","sentAt":"2009-11-24T23:08:12Z","receivedAt":"2009-11-24T23:08:12Z","isPatch":false,"sender":{"key":"nanako3@lavabit.com","avatar":"https://gravatar.com/avatar/3777b9e201c5883a62b1a6fdf7c53f2d712d1d80989146063ea861e33aad72a8?d=mp&s=160"},"body":"Quoting Avery Pennarun \n\n> Yup. This is basically a bug in \"git merge -s subtree\": it guesses\n> which subtree to merge into, rather than actually taking a prefix\n> parameter. I've been meaning to either submit a patch for this, or\n> find a way to work around it.\n\nProbably you can save time by using what was already done\n\nhttp://thread.gmane.org/gmane.comp.version-control.git/76650/focus=89021\n\n\n\n-- \nNanako Shiraishi\nhttp://ivory.ap.teacup.com/nanako3/\n"},{"id":"128313","messageId":"20091125143204.GA9834@lonquimay.wrk.lsn.camptocamp.com","threadId":"21741","inReplyTo":"32541b130911241348s21e21fb8n12edf374e6a3c309@mail.gmail.com","subject":"Re: git-subtree: directory mismatch","fromName":"Marc Fournier","fromEmail":"marc.fournier@camptocamp.com","sentAt":"2009-11-25T14:32:04Z","receivedAt":"2009-11-25T14:32:04Z","isPatch":false,"sender":{"key":"marc.fournier@camptocamp.com","avatar":null},"body":"On Tue, Nov 24, 2009 at 04:48:26PM -0500, Avery Pennarun wrote:\n> On Tue, Nov 24, 2009 at 2:53 PM, Marc Fournier\n> wrote:\n> > I was not able to reproduce this bug in any of the following cases:\n> >  - if the README file wasn't empty\n> >  - if modules/cognac is renamed to something else\n> >  - if modules/cognac doesn't have the same sub-directories than\n> >   modules/mapserver\n> >\n> > This bug happens using git 1.5.5.6 as well as 1.6.5. It seems to happen\n> > when git-subtree calls \"git merge -s subtree\".\n> \n> Yup. This is basically a bug in \"git merge -s subtree\": it guesses\n> which subtree to merge into, rather than actually taking a prefix\n> parameter. I've been meaning to either submit a patch for this, or\n> find a way to work around it.\n\nI see. What sort of criteria does git-merge use to do this auto-guessing ?\n\nI'm hoping to find a way to workaround this problem. I have a couple of\nsubtrees I'm completely unable to update. The best I found is:\n\"git rm -fr ...\" followed by \"git subtree add ...\" again.\n\n> This doesn't usually happen once your project is relatively mature\n> (ie. doesn't have blank or \"default\" template files in it) since then\n> the auto-guessing gets more reliable. But there's no good reason to\n> do the auto-guessing, so it would be best to do this \"properly.\"\n\nThe thing is I use dozens of different \"modules\" in my project, all merged\ninto using git-subtree. All these modules have the same structure, and most\nof them have quite minimal content and won't really evolve more. In my\ncase, evolution will mean adding even more modules.\n\nIf I understand the thread Nanako Shiraishi pointed us to, a patch\nimplementing a prefix to \"git merge -s subtree\" is available somewhere but\nhasn't been added in any official release yet ?\n\nMarc\n"},{"id":"128326","messageId":"32541b130911251028h6db240d5yd171fa4941ef14ba@mail.gmail.com","threadId":"21741","inReplyTo":"20091125080812.6117@nanako3.lavabit.com","subject":"Re: git-subtree: directory mismatch","fromName":"Avery Pennarun","fromEmail":"apenwarr@gmail.com","sentAt":"2009-11-25T18:28:52Z","receivedAt":"2009-11-25T18:28:52Z","isPatch":false,"sender":{"key":"apenwarr@gmail.com","avatar":"https://avatars.githubusercontent.com/u/20592?v=4"},"body":"On Tue, Nov 24, 2009 at 6:08 PM, Nanako Shiraishi wrote:\n> Quoting Avery Pennarun \n>\n>> Yup.  This is basically a bug in \"git merge -s subtree\": it guesses\n>> which subtree to merge into, rather than actually taking a prefix\n>> parameter.  I've been meaning to either submit a patch for this, or\n>> find a way to work around it.\n>\n> Probably you can save time by using what was already done\n>\n> http://thread.gmane.org/gmane.comp.version-control.git/76650/focus=89021\n\nHi Nanako,\n\nI've read through the thread (I do remember skimming it awhile ago)\nbut can't find patches for the syntax actually under discussion. I\nfound a patch that introduces \"-s theirs\", which was summarily shot\ndown in favour of adding -X support, but I can't find any actual patch\nfor this. Moreover, git-merge seems to have been ported to C since\nthen, so I guess it wouldn't apply anyway. And I can't find any\nimplementation at all for the discussed \"-Xsubtree=\" option, which I'm\nguessing didn't actually ever get done.\n\nAm I missing something? It looks fairly easy to throw in anyway, but\nit's even easier if someone has already done it :)\n\nThanks,\n\nAvery\n"},{"id":"128328","messageId":"7v7htexuxc.fsf@alter.siamese.dyndns.org","threadId":"21741","inReplyTo":"32541b130911251028h6db240d5yd171fa4941ef14ba@mail.gmail.com","subject":"Re: git-subtree: directory mismatch","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2009-11-25T19:31:59Z","receivedAt":"2009-11-25T19:31:59Z","isPatch":false,"sender":{"key":"gitster@pobox.com","avatar":"https://avatars.githubusercontent.com/u/54884?v=4"},"body":"Avery Pennarun writes:\n\n>> Probably you can save time by using what was already done\n>>\n>> http://thread.gmane.org/gmane.comp.version-control.git/76650/focus=89021\n>\n> Hi Nanako,\n>\n> I've read through the thread (I do remember skimming it awhile ago)\n> but can't find patches for the syntax actually under discussion.\n\nI very much prefer gmane threading when following discussion over all the\nother mail archives, but this shows one thing I really dislike about it.\n\nIt is not easy to find a near-by thread when looking at an old article,\nand you have to be willing to bisect the page number at the right hand\nside of the web UI. Often a patch series is posted as a separate thread\nafter a discussion reaches conclusion or identifies an issue to solve, and\nthe real patch series lives in a near-by thread. Very inconvenient.\n\nI don't know how Nana digs up older discussion; maybe she knows better\nways.\n\nIn my primary repository, I have an archive of mothballed branches kept\nwith this alias:\n\n [alias]\n hold = \"!sh -c 'git update-ref refs/hold/$1 refs/heads/$1 && git branch -D $1' -\"\n\nand found this series in there. It applies to v1.6.0-rc0~245 (no, I won't\nbe rebasing this myself---I don't have time for that while preparing for\nthe pre-release feature freeze).\n\n f7713ce Document that merge strategies can now take their own options\n 7eda236 Make \"subtree\" part more orthogonal to the rest of merge-recursive.\n e416d61 Teach git-pull to pass -X