How do I force "git pull" to overwrite local files?

5 544

2 192

How do I force an overwrite of local files on a git pull?

The scenario is following:

  • A team member is modifying the templates for a website we are working on
  • They are adding some images to the images directory (but forgets to add them under source control)
  • They are sending the images by mail, later, to me
  • I'm adding the images under the source control and pushing them to GitHub together with other changes
  • They cannot pull updates from GitHub because Git doesn't want to overwrite their files.

This is the error I'm getting:

error: Untracked working tree file 'public/images/icon.gif' would be overwritten by merge

How do I force Git to overwrite them? The person is a designer - usually I resolve all the conflicts by hand, so the server has the most recent version that they just needs to update on their computer.

Jakub Troszok

Posted 2009-07-14T14:58:15.550

Reputation: 30 961

6anyone reading this who thinks they might lose files, I've been in this position and found Sublime Text's buffer has saved me - if I'm working on something, then accidentally delete everything by trying to solve a similar problem to this or by using an answer on this question and have had the files open in Sublime (which there's a good chance of) then the files will still be there is Sublime, either just there, or in the undo history – Toni Leigh – 2016-01-20T08:51:22.953

6git reset --hard origin/branch_to_overwrite – Andrew Atkinson – 2016-03-22T08:37:48.723

basically, only do a pull from develop after the initial checkout -b. do your work, then push back in. – ldgorman – 2018-08-22T09:09:33.347

Short answer: delete and re-create branch.

  1. Delete branch: git branch <branch> -D
  2. Reset to a commit before the conflict: git reset <commit> --hard
  3. Re-create the branch: git branch <branch>
  4. Set tracking to the server: `git --set-upstream-to=origin/<branch> <branch>
  5. Pull: git pull
  6. < – Nino Filiu – 2018-09-24T08:54:52.033

Answers

7 737

Important: If you have any local changes, they will be lost. With or without --hard option, any local commits that haven't been pushed will be lost.[*]

If you have any files that are not tracked by Git (e.g. uploaded user content), these files will not be affected.


I think this is the right way:

git fetch --all

Then, you have two options:

git reset --hard origin/master

OR If you are on some other branch:

git reset --hard origin/<branch_name>

Explanation:

git fetch downloads the latest from remote without trying to merge or rebase anything.

Then the git reset resets the master branch to what you just fetched. The --hard option changes all the files in your working tree to match the files in origin/master


Maintain current local commits

[*]: It's worth noting that it is possible to maintain current local commits by creating a branch from master before resetting:

git checkout master
git branch new-branch-to-save-current-commits
git fetch --all
git reset --hard origin/master

After this, all of the old commits will be kept in new-branch-to-save-current-commits.

Uncommitted changes

Uncommitted changes, however (even staged), will be lost. Make sure to stash and commit anything you need. For that you can run the following:

git stash

And then to reapply these uncommitted changes:

git stash pop

RNA

Posted 2009-07-14T14:58:15.550

Reputation: 83 410

3Watch out! If you have local unpushed commits this will remove them from your branch! This solution keeps untracked files not in the repository intact, but overwrites everything else. – Matthijs P – 2012-05-17T08:18:25.470

413It's a popular question, so I'd like to clarify on the top comment here. I just executed commands as described in this answer and it hasn't removed ALL the local files. Only the remotely tracked files were overwritten, and every local file that has been here was left untouched. – Red – 2012-11-22T10:38:05.530

11This worked for me and my local files were NOT deleted. – Tastybrownies – 2013-05-21T18:16:11.237

1I lost all local commits that weren't in the origin/master tree after a rebase. It's very possible to shoot yourself in the foot, here. – Cuadue – 2013-11-22T22:22:07.227

You can also do this for a tracking branch if you have the same branch in origin and locally, just change git reset --hard origin/master to git reset --hard origin/(branch) – Joshua Kolden – 2013-12-14T05:59:12.140

10in case you're pulling from a repo that has its remote branch name different from "master", use git reset --hard origin/branch-name – Nerrve – 2013-12-17T11:17:24.430

This should be the accepted answer. Works fine for me. I guess the assumption here though is that you do not have local commits, which in my case is good as the "local" only possesses a deployment key. – Ardee Aram – 2014-04-07T04:02:37.437

Won't this also throw away my local commits? I want to keep my commits, and only overwrite untracked/uncommitted changes. – mcv – 2014-05-02T15:15:24.600

2fetch origin master may be more appropriate if you just want one branch... – Jorge Leitão – 2014-05-19T07:56:30.183

I just use "git fetch" to get the upstream branch and then reset when I need to fix a bunch of files. You can get the upstream branch with "git branch -vv" – nathanengineer – 2014-08-17T23:32:15.030

1@mcv To keep your local commits you want git reset --hard HEAD which you could then follow by git merge origin/master. – joeytwiddle – 2014-08-19T03:47:08.497

19Given the amount of upvotes to this question and answer, I think that git should incorporate a command like git pull -f – Sophivorus – 2014-08-26T01:33:48.960

3Commits that weren't pushes before the hard reset can be recovered using git reflog, which list all commits, also those without a base. Until you cleanup your local copy using git gc, then all is lost – Koen. – 2015-02-10T22:24:27.473

Will this delete files that I have .gitignore'd? – user3817250 – 2015-04-01T14:27:29.070

2@FelipeSchenone There actually is git pull -f, but it doesn't do what we all want it to do. – sudo – 2015-05-01T07:49:21.387

3To be correct, the uncommited changes are lost but old commits are dangling, waiting for garbage collection. You can still see them in git reflog and move your HEAD back with git reset --hard [email protected]{1}. – Florian Fida – 2015-05-29T03:01:36.293

2Want to point out here that it will also remove any changes made in any other branch which is not tracked remotely. I lost mine, hope others be careful. – Devesh Khandelwal – 2015-06-27T16:35:19.203

2Backup the Folder with your local files including the .git folder in it. Now you can't 'loose' anything if this fix doesn't work for you; you can make another copy and try something else. Do not operate out of a single folder and trust git as if it were a 100% reliable backup tool. One typo, especially in branch-land, can be a nightmare. I suggest one backup old-school to tar.gz folders, and then run git out of the latest copy. – JosephK – 2015-08-13T18:44:38.787

2@Red Yep, if you want to remove everything untracked, even ignored, use git clean -dxf after resetting hard. – IceGlow – 2015-12-31T08:52:02.953

I did this and my local changes were NOT discarded - which is unfortunate because I was hoping they would be! Does anyone know how to do this, but also force git to discard any locale changes that don't exist in the remote branch? I essentially want an EXACT, 1-to-1 correspondence with the remote branch. – Gershom Maes – 2016-01-04T15:53:50.767

2Little known fact - you can actually back up your whole git repo with good old fashioned tar -czf myrepodir.tar.gz myrepodir/.. If your repo gets all messed up due to git weirdness, delete it and restore from the tar.gz file to ACTUALLY get back to where you were before you tried this ill-fated experiment. Just make sure you have the . after myrepodir/ or your .git and other hidden files won't be saved! – Buttle Butkus – 2016-02-06T08:19:17.523

can this technique work for individual files? for example -- index.php helpers.js ect. in place of --all ? – Steve C – 2016-09-09T15:47:54.137

didnt work for me. tried this solution, tried git clean -dfx. tried git pull -f. still have the same error message. – InsOp – 2016-10-04T09:47:35.610

1Note that you can abbreviate the branch name by @{u}, which is much shorter. – Antimony – 2017-06-06T17:56:04.297

@GershomMaes git clean -dfx after hard reset, as mentioned above by IceGlow?

– Franklin Yu – 2018-03-23T19:40:57.323

isn't it easier just to rm -rf /folder and do git clone again, if you want to discard all your local changes? – JonB – 2018-10-04T14:24:59.040

787

Try this:

git reset --hard HEAD
git pull

It should do what you want.

Travis Reeder

Posted 2009-07-14T14:58:15.550

Reputation: 22 042

232instead of merging using 'git pull', try git fetch --all followed by 'git reset --hard origin/master' – Lloyd Moore – 2012-02-21T14:56:54.207

Lloyd Moore's suggestion is correct, but beware that it can remove local unpushed commits from your branch. – Matthijs P – 2012-05-17T08:23:04.497

5yep, the @lloydmoore solution worked for me. Could do with being an answer rather than just a comment. – Max Williams – 2012-11-19T09:54:51.117

2This will reset the current changes back to the last branch commit pulled. Then git pull merges the changes from the latest branch. This did exactly what I wanted it to do.. Thanks! – Codeversed – 2014-12-05T17:42:09.430

Will this mess the remote? – user1767754 – 2015-09-30T18:33:51.850

This will merge, so it won't overwrite as the OP wanted... – Luis Ortega Araneda – 2016-01-21T14:17:46.410

12I've done this and some local files that were no longer in repo were left on the disk. – Piotr Owsiak – 2011-04-08T16:00:57.413

1Good answer, but please add more detail, I mean some people may don't know anything about the reset and lose their files... – M98 – 2017-01-17T12:54:17.103

@Travis R could you please explain the purpose of git rest --hard HEAD ? – Imanez – 2017-03-06T15:03:04.953

19I do not think that this is correct. the above will perform a merge, not overwrite which was requested in the question: "How to force git to overwrite them?" I do not have the answer, I am currently looking for it.. at the moment I switch to the branch with with the code that I want to keep "git checkout BranchWithCodeToKeep", then do "git branch -D BranchToOverwrite" and then finally "git checkout -b BranchToOverwrite". you will now have the exact code from BranchWithCodeToKeep on the branch BranchToOverwrite without having to perform a merge. – felbus – 2011-07-13T10:11:31.767

git reset --hard HEAD doesn't really reset all in some cases – Finesse – 2018-03-06T02:19:35.513

This is mostly what I think of when I think of forcing the pull. Mostly after I tried some settings – Mianala Loharano – 2018-07-04T09:17:41.843

400

WARNING: git clean deletes all your untracked files/directories and can't be undone.


Sometimes just clean -f does not help. In case you have untracked DIRECTORIES, -d option also needed:

# WARNING: this can't be undone!

git reset --hard HEAD
git clean -f -d
git pull

WARNING: git clean deletes all your untracked files/directories and can't be undone.

Consider using -n (--dry-run) flag first. This will show you what will be deleted without actually deleting anything:

git clean -n -f -d

Example output:

Would remove untracked-file-1.txt
Would remove untracked-file-2.txt
Would remove untracked/folder
...

David Avsajanishvili

Posted 2009-07-14T14:58:15.550

Reputation: 5 254

29Awesome... Ran this against my dotfiles repo... In my home directory. Good that I didn't really have anything important there... – Lauri – 2011-12-11T10:35:13.573

5I think the scenario description makes it clear that he doesn't really want to throw away the content. Rather what he wants is to stop git baulking at overwriting the files. @Lauri, this should not have happened to you. Unfortunately people seem to have misread the essence of scenario description - see my suggestion. – Hedgehog – 2012-02-11T23:05:41.997

14FINALLY. git clean -f -d is handy when make clean fails to clean everything. – earthmeLon – 2012-06-23T04:32:08.750

see https://www.kernel.org/pub/software/scm/git/docs/v1.7.2.2/git-clean.html

– Peter Ehrlich – 2012-07-19T16:58:54.687

If you want to sync the whole directory with the remote repo, this is the way to go. – Johannes Ewald – 2013-02-11T12:04:55.560

7@crizCraig unless they are added in .gitignore – Bleeding Fingers – 2013-06-13T06:58:23.683

@BleedingFingers Then add -x. But existing files that are covered by .gitignore shouldn't be a problem: assuming the other contributors to that repo use the same .gitignore file, a pull won't receive any commits that add conflicting files. – None – 2015-06-07T00:18:43.007

3@earthmeLon, for that you might want git clean -dfx. The -x ignores .gitignore. Typically your build products will be in .gitignore. – Paul Draper – 2015-08-12T18:28:00.420

1It can be good to run git clean -nfd first, which only shows, what would be removed before it's actually done, that is before you run into troubles. :-) – Velda – 2018-08-16T14:26:41.227

You can give git clean a path argument to be more specific and avoid deleting untracked files that aren't conflicting. – joachim – 2011-10-18T10:08:13.190

344

Like Hedgehog I think the answers are terrible. But though Hedgehog's answer might be better, I don't think it is as elegant as it could be. The way I found to do this is by using "fetch" and "merge" with a defined strategy. Which should make it so that your local changes are preserved as long as they are not one of the files that you are trying to force an overwrite with.

First do a commit of your changes

 git add *
 git commit -a -m "local file server commit message"

Then fetch the changes and overwrite if there is a conflict

 git fetch origin master
 git merge -s recursive -X theirs origin/master

"-X" is an option name, and "theirs" is the value for that option. You're choosing to use "their" changes, instead of "your" changes if there is a conflict.

Richard Kersey

Posted 2009-07-14T14:58:15.550

Reputation: 3 673

53This is the best answer I've seen so far. I haven't tried it, but unlike other answers, this doesn't attempt to nuke all your untracked files, which is very dangerous for obvious reasons. – huyz – 2012-05-07T09:36:16.383

4Ditto - this worked for me when doing a very large merge (GitHub pull request) where I just wanted to accept it all on top of what I had. Good answer! In my case the last two commands were: 1) get fetch other-repo; 2) git merge -s recursive -X theirs other-repo/master – quux00 – 2012-07-27T01:44:56.333

4Question: What is -X? What is 'theirs'? – AlxVallejo – 2014-11-21T14:50:07.853

2This will overwrite any conflicts with the repositories files and not your local ones, correct? – Nathan Fiscaletti – 2014-12-05T11:40:52.770

5"-X" is an option name, and "theirs" is the value for that option. You're choosing to use "their" changes, instead of "your" changes if there is a conflict. – Richard Kersey – 2015-08-21T21:51:23.287

1Best answer. The highest accepted answer left me in my case on detached head. I switched back to local master branch and ran git merge -X theirs origin/master – petergus – 2016-03-11T12:46:21.710

Tried this - then add/commit my new changed files, but still said my branch was behind remote when I tried to push. Massive wasted time because "git pull" doesn't ask "Overwrite local? y/n/all". – JosephK – 2016-07-26T03:57:11.887

1Hang on...doesn't have git add * and git commit -a &lt;more-options-here&gt; have the same effect? Why would you need both? – Marcel Stör – 2017-04-25T20:41:26.190

@MarcelStör you don't need both but they also don't do the same thing. git commit -a only has that affect on files that are already tracked; git add * will add files that are not tracked. So if you do git commit -a it won't catch new files, then you might need to do a git add afterward, but it you do git add * you won't need -a on the commit. – briantist – 2018-04-04T20:47:14.060

1The problem with this (excellent) answer, is it adds the all the local files, which sometimes may not be what you want. You may just want to add the specific files that were omitted.

But the best thing about it is, it gets him to do what he should have done -- add them locally.

You probably won't need the -X theirs strategy, since they're the same image. In fact, I'd suggest leaving it off at first, just to find out if there are any anomalies, and add it in if there are, after reviewing that 'theirs' is always the correct choice.

But then, I'm paranoid. – Bob Kerns – 2018-06-28T20:27:02.897

242

Instead of doing:

git fetch --all
git reset --hard origin/master

I'd advise doing the following:

git fetch origin master
git reset --hard origin/master

No need to fetch all remotes and branches if you're going to reset to the origin/master branch right?

Johanneke

Posted 2009-07-14T14:58:15.550

Reputation: 3 135

3Your answer is just what you needed for your rep. I must ask, does this also remove all untracked files? – Nicolas De Jay – 2014-01-07T06:38:09.083

5Yeah, most of my rep is coming from here :) This will also remove all untracked files. Something I had forgotten and was painfully reminded of just 2 days ago... – Johanneke – 2014-01-09T12:01:15.310

1

See the comments on this other answer: http://stackoverflow.com/a/8888015/2151700

– Johanneke – 2014-01-09T12:02:55.517

This did not remove my untracked files; which is actually what I'd expect. Is there a reason it might for some people and not for others? – arichards – 2016-04-19T15:27:08.763

Untracked files are not affect4ed by git reset. If you want them to be removed as well, do git add . first, before git reset --hard – Johanneke – 2017-08-15T09:12:21.860

This is exactly what I needed: something that overwrites untracked files that exist in the remote, and leaves everything else intact. – Ledazinha – 2017-12-20T22:37:58.697

124

It looks like the best way is to first do:

git clean

To delete all untracked files and then continue with the usual git pull...

Jakub Troszok

Posted 2009-07-14T14:58:15.550

Reputation: 30 961

I needed to add a -x to that git clean to get it to work for me, for some reason (-d wasn't deleting a .ideas directory for some reason), but this fixed my problem, certainly. – Owen Blacker – 2011-12-19T12:29:27.883

3I tried using "git clean" to solve the same issue, but it did not resolve it.

git status says "Your branch and 'origin/master' have diverged,

and have 2 and 9 different commit(s) each, respectively." and git pull says something similar to what you have above.< – slacy – 2009-09-24T04:25:42.083

would he not need to change to the master branch and then do git clean for it to work? Then he can flip back to whatever development branch he was working from. – suitedupgeek – 2010-01-25T21:38:00.687

@slacy: Then you need to merge the two branches. – Xiong Chiamiov – 2010-02-05T00:07:29.003

avoid sudden cardiac arrest: use git clean -n first – doub1ejack – 2013-07-30T14:19:40.240

36git clean is a rather blunt instrument, and could throw away a lot of things that you may want to keep. Better to remove or rename the files that git is complaining about until the pull succeeds. – Neil Mayhew – 2010-07-02T13:21:35.287

2Is git clean the best answer here? Seems like removing files isn't necessarily what the OP wants. They asked for 'an overwrite of local files' not deletion. – JohnAllen – 2014-03-04T08:28:11.683

2I do not think this works in general. Isn't there a way to do basically a git clone remote via a forced git pull? – mathtick – 2010-11-29T18:30:01.267

1git clean by itself doesn't seem to help, @Arrowmaster 's solution is closer, but a better one by Lloyd Moore is in the next answer. – Neil Monroe – 2015-06-10T21:12:20.357

9@mathick: git fetch origin &amp;&amp; git reset --hard origin/master – Arrowmaster – 2011-02-23T04:24:56.047

Don't ever do anything 'git' (or even have a .git folder) except in a copy of your actual work folder, and avoid all cardiac stress. Don't give git and its obtuse syntax/methods the opportunity to wreck you in the first place. – JosephK – 2016-07-26T04:12:59.867

100

Warning, doing this will permanently delete your files if you have any directory/* entries in your gitignore file.

Some answers seem to be terrible. Terrible in the sense of what happened to @Lauri by following David Avsajanishvili suggestion.

Rather (git > v1.7.6):

git stash --include-untracked
git pull

Later you can clean the stash history.

Manually, one-by-one:

$ git stash list
[email protected]{0}: WIP on <branch>: ...
[email protected]{1}: WIP on <branch>: ...

$ git stash drop [email protected]{0}
$ git stash drop [email protected]{1}

Brutally, all-at-once:

$ git stash clear

Of course if you want to go back to what you stashed:

$ git stash list
...
$ git stash apply [email protected]{5}

Hedgehog

Posted 2009-07-14T14:58:15.550

Reputation: 3 367

Aren't you assuming that a commit has never been performed? In my case I made several commits to my local branch, but wanted to "reset" everything back to the remote branch – Lee Francis – 2012-03-20T10:55:38.690

1No I don't think so. Stashing just moves uncommitted files out of the way. The above also moves (stashes) files that git does not track. This prevents files that have been added to the remote, which have not yet pulled down to your machine - but which you have created (!) - to be pulled down. All without destroying the uncommitted work. Hope that makes sense? – Hedgehog – 2012-03-20T23:54:34.997

3If you don't have 1.7.6, you can mimic --include-untracked simply by temporarily git add-ing your entire repo, then immediately stashing it. – nategood – 2012-05-01T22:48:15.530

2I agree with Hedgehog. If you do the popular answers here, you are more than likely going to find you've inadvertently killed a lot of stuff that you didn't really want to lose. – Guardius – 2013-01-31T21:28:03.027

1I had other untracked files--besides the one the merge/pull wanted to overwrite, so this solution worked best. git stash apply brought back all my untracked files with the exception (rightly) of the ones that the merge had already created: "already exists, no checkout." Worked perfectly. – BigBlueHat – 2013-04-25T04:55:09.853

This can be the easiest way to hose a repo unintentionally, just like what happened to @Lauri, but worse because you think you're protecting files from deletion. If you have one .gitignore rule that has a wildcard in it, kiss them goodbye. Explanation.

– Walf – 2016-11-30T07:05:22.403

2This is the cleanest answer, and should be the accepted one. To save some typing you can use the short form: git stash -u. – ccpizza – 2017-03-23T08:30:32.740

86

You might find this command helpful to throw away local changes:

git checkout <your-branch> -f

And then do a cleanup (removes untracked files from the working tree):

git clean -f

If you want to remove untracked directories in addition to untracked files:

git clean -fd

Vishal

Posted 2009-07-14T14:58:15.550

Reputation: 10 677

I think the scenario description makes it clear that he doesn't really want to throw away the content. Rather what he wants is to stop git baulking at overwriting the files. See my suggestion. – Hedgehog – 2012-02-11T23:03:53.327

2Though that answer might not fit exactly the description, it still saved me from the frustration of git twiddling with the carriage returns (event with autocrlf false). When git reset --hard HEAD does not leave you with "no" modified files, these "-f" flags are quite helpful. Thanks a bunch. – Kellindil – 2013-01-16T10:28:58.430

78

Instead of merging with git pull, try this:

git fetch --all

followed by:

git reset --hard origin/master.

Lloyd Moore

Posted 2009-07-14T14:58:15.550

Reputation: 2 515

53

The only thing that worked for me was:

git reset --hard HEAD~5

This will take you back five commits and then with

git pull

I found that by looking up how to undo a Git merge.

Chris BIllante

Posted 2009-07-14T14:58:15.550

Reputation: 563

This was what ultimately worked for me as I had force pushed my branch to the origin repo and kept getting merge conflicts when trying to pull it to my remote repo.. – jwfrench – 2014-05-07T05:16:17.253

Hi, actually this is a trick for a work around but really effective. Because some conflicts may happen just in few commits then reverting 5 commits will make sure no conflicts with remote code. – Hoang Le – 2014-11-21T10:03:43.023

50

The problem with all these solutions is that they are all either too complex, or, an even bigger problem, is that they remove all untracked files from the web server, which we don't want since there are always needed configuration files which are on the server and not in the Git repository.

Here is the cleanest solution which we are using:

# Fetch the newest code
git fetch

# Delete all files which are being added, so there
# are no conflicts with untracked files
for file in `git diff HEAD..origin/master --name-status | awk '/^A/ {print $2}'`
do
    rm -f -- "$file"
done

# Checkout all files which were locally modified
for file in `git diff --name-status | awk '/^[CDMRTUX]/ {print $2}'`
do
    git checkout -- "$file"
done

# Finally pull all the changes
# (you could merge as well e.g. 'merge origin/master')
git pull
  • The first command fetches newest data.

  • The second command checks if there are any files which are being added to the repository and deletes those untracked files from the local repository which would cause conflicts.

  • The third command checks-out all the files which were locally modified.

  • Finally we do a pull to update to the newest version, but this time without any conflicts, since untracked files which are in the repo don't exist anymore and all the locally modified files are already the same as in the repository.

Strahinja Kustudic

Posted 2009-07-14T14:58:15.550

Reputation: 2 347

Using "git merge origin/master" as the last line (like you say in your note) instead of "git pull" will be faster as you've already pulled down any changes from the git repo. – Josh – 2013-05-06T06:21:24.097

1Yeah of course, git merge origin/master will be faster and probably even safer. Since if someone pushed new changes during the removal of of files of this script (which is not likely to happen, but possible), the whole pull could fail.

The only reason I put pull in there is because someone might not be working on the master branch, but some other branch and I wanted the script to be universal. – Strahinja Kustudic – 2013-09-01T22:25:37.063

If you have locally created files like option files, put them in .gitignore. – Sebi – 2017-11-21T11:41:39.653

38

I had the same problem. No one gave me this solution, but it worked for me.

I solved it by:

  1. Deleting all the files. Leave just the .git directory.
  2. git reset --hard HEAD
  3. git pull
  4. git push

Now it works.

John John Pichler

Posted 2009-07-14T14:58:15.550

Reputation: 2 473

1Same here. Sometimes only the very hard solution works, it happens often that only reset and clean are not enough somehow... – jdehaan – 2011-12-15T11:28:09.653

32

First of all, try the standard way:

git reset HEAD --hard # Remove all not committed changes

If above won't help and you don't care about your untracked files/directories (make the backup first just in case), try the following simple steps:

cd your_git_repo  # where 'your_git_repo' is your git repository folder
rm -rfv *         # WARNING: only run inside your git repository!
git pull          # pull the sources again

This will REMOVE all git files (excempt .git/ dir, where you have all commits) and pull it again.


Why git reset HEAD --hard could fail in some cases?

  1. Custom rules in .gitattributes file

    Having eol=lf rule in .gitattributes could cause git to modify some file changes by converting CRLF line-endings into LF in some text files.

    If that's the case, you've to commit these CRLF/LF changes (by reviewing them in git status), or try: git config core.autcrlf false to temporary ignore them.

  2. File system incompability

    When you're using file-system which doesn't support permission attributes. In example you have two repositories, one on Linux/Mac (ext3/hfs+) and another one on FAT32/NTFS based file-system.

    As you notice, there are two different kind of file systems, so the one which doesn't support Unix permissions basically can't reset file permissions on system which doesn't support that kind of permissions, so no matter how --hard you try, git always detect some "changes".

kenorb

Posted 2009-07-14T14:58:15.550

Reputation: 63 134

27

Bonus:

In speaking of pull/fetch/merge in the previous answers, I would like to share an interesting and productive trick,

git pull --rebase

This above command is the most useful command in my Git life which saved a lot of time.

Before pushing your newly commit to server, try this command and it will automatically synchronise the latest server changes (with a fetch + merge) and will place your commit at the top in the Git log. There isn't any need to worry about manual pull/merge.

Find details in What does "git pull --rebase" do?.

Sazzad Hissain Khan

Posted 2009-07-14T14:58:15.550

Reputation: 14 113

27

I had a similar problem. I had to do this:

git reset --hard HEAD
git clean -f
git pull

Ryan

Posted 2009-07-14T14:58:15.550

Reputation: 311

4use git clean with caution – nategood – 2012-03-30T16:39:22.187

26

I summarized other answers. You can execute git pull without errors:

git fetch --all
git reset --hard origin/master
git reset --hard HEAD
git clean -f -d
git pull

Warning: This script is very powerful, so you could lose your changes.

Robert Moon

Posted 2009-07-14T14:58:15.550

Reputation: 762

2This will overwrite modified files (files that were previously checked in) and it will remove untracked files (files that have never been checked in). Exactly what I was looking for, thanks! – styfle – 2016-03-03T16:01:46.527

3I suspect the third line git reset --hard HEAD may be redundant; my local man page (2.6.3) say that reset in the second line git reset --hard origin/master "defaults to HEAD in all forms." – arichards – 2016-04-19T15:40:52.043

2@arichards I think your suspect is right but if second line will not work(by any reason) third line work well to reset. This solution doesn't need to be optimized. I just summarized other answers. That's all. Thank you for your comment. :) – Robert Moon – 2016-04-20T02:12:45.603

25

Based on my own similar experiences, the solution offered by Strahinja Kustudic above is by far the best. As others have pointed out, simply doing hard reset will remove all the untracked files which could include lots of things that you don't want removed, such as config files. What is safer, is to remove only the files that are about to be added, and for that matter, you'd likely also want to checkout any locally-modified files that are about to be updated.

That in mind, I updated Kustudic's script to do just that. I also fixed a typo (a missing ' in the original).

#/bin/sh

# Fetch the newest code
git fetch

# Delete all files which are being added,
# so there are no conflicts with untracked files
for file in `git diff HEAD..origin/master --name-status | awk '/^A/ {print $2}'`
do
    echo "Deleting untracked file $file..."
    rm -vf "$file"
done

# Checkout all files which have been locally modified
for file in `git diff HEAD..origin/master --name-status | awk '/^M/ {print $2}'`
do
    echo "Checking out modified file $file..."
    git checkout $file
done

# Finally merge all the changes (you could use merge here as well)
git pull

Rolf Kaiser

Posted 2009-07-14T14:58:15.550

Reputation: 439

Using "git merge origin/master" as the last line (like you say in your note) instead of "git pull" will be faster as you've already pulled down any changes from the git repo. – Josh – 2013-05-06T06:20:47.517

The checkout of modified files is needed, so this works 100% of times. I updated my script with that a long time ago, but forgot to update here as well. I also use it a little differently than you. I checkout files which have any type of modification, not just M, so it works all the time. – Strahinja Kustudic – 2013-09-01T22:48:40.013

22

I believe there are two possible causes of conflict, which must be solved separately, and as far as I can tell none of the above answers deals with both:

  • Local files that are untracked need to be deleted, either manually (safer) or as suggested in other answers, by git clean -f -d

  • Local commits that are not on the remote branch need to be deleted as well. IMO the easiest way to achieve this is with: git reset --hard origin/master (replace 'master' by whatever branch you are working on, and run a git fetch origin first)

tiho

Posted 2009-07-14T14:58:15.550

Reputation: 4 307

20

An easier way would be to:

git checkout --theirs /path/to/file.extension
git pull origin master

This will override your local file with the file on git

maximus 69

Posted 2009-07-14T14:58:15.550

Reputation: 858

19

It seems like most answers here are focused on the master branch; however, there are times when I'm working on the same feature branch in two different places and I want a rebase in one to be reflected in the other without a lot of jumping through hoops.

Based on a combination of RNA's answer and torek's answer to a similar question, I've come up with this which works splendidly:

git fetch
git reset --hard @{u}

Run this from a branch and it'll only reset your local branch to the upstream version.

This can be nicely put into a git alias (git forcepull) as well:

git config alias.forcepull "!git fetch ; git reset --hard @{u}"

Or, in your .gitconfig file:

[alias]
  forcepull = "!git fetch ; git reset --hard @{u}"

Enjoy!

JacobEvelyn

Posted 2009-07-14T14:58:15.550

Reputation: 2 339

This answer is also nice because it works regardless of which branch you are on! – leafmeal – 2018-09-11T20:14:56.590

18

I just solved this myself by:

git checkout -b tmp # "tmp" or pick a better name for your local changes branch
git add -A
git commit -m 'tmp'
git pull
git checkout master # Or whatever branch you were on originally
git pull
git diff tmp

where the last command gives a list of what your local changes were. Keep modifying the "tmp" branch until it is acceptable and then merge back onto master with:

git checkout master && git merge tmp

For next time, you can probably handle this in a cleaner way by looking up "git stash branch" though stash is likely to cause you trouble on the first few tries, so do first experiment on a non-critical project...

Simon B.

Posted 2009-07-14T14:58:15.550

Reputation: 1 256

18

I had the same problem and for some reason, even a git clean -f -d would not do it. Here is why: For some reason, if your file is ignored by Git (via a .gitignore entry, I assume), it still bothers about overwriting this with a later pull, but a clean will not remove it, unless you add -x.

Tierlieb

Posted 2009-07-14T14:58:15.550

Reputation: 227

15

I have a strange situation that neither git clean or git reset works. I have to remove the conflicting file from git index by using the following script on every untracked file:

git rm [file]

Then I am able to pull just fine.

Chen Zhang

Posted 2009-07-14T14:58:15.550

Reputation: 167

13

These four commands work for me.

git reset --hard HEAD
git checkout origin/master
git branch -D master
git checkout -b master

To check/pull after executing these commands

git pull origin master

I tried a lot but finally got success with these commands.

vishesh chandra

Posted 2009-07-14T14:58:15.550

Reputation: 5 540

2"git branch -D master" delete the branch. so be careful with it. I prefer to use "git checkout origin/master -b <new branch name>" which create a new branch with a new name and you done need 3,4 lines. Also recommended to use "git clean -f" as well. – Chand Priyankara – 2014-04-05T11:49:57.130

12

Despite the original question, the top answers can cause problems for people who have a similar problem, but don't want to lose their local files. For example, see Al-Punk and crizCraig's comments.

The following version commits your local changes to a temporary branch (tmp), checks out the original branch (which I'm assuming is master) and merges the updates. You could do this with stash, but I've found it's usually easier to simply use the branch / merge approach.

git checkout -b tmp
git add *; git commit -am "my temporary files"
git checkout master

git fetch origin master
git merge -s recursive -X theirs origin master

where we assume the other repository is origin master.

Snowcrash

Posted 2009-07-14T14:58:15.550

Reputation: 36 593

12

Just do

git fetch origin branchname
git checkout -f origin/branchname // This will overwrite ONLY new included files
git checkout branchname
git merge origin/branchname

So you avoid all unwanted side effects, like deleting files or directories you wanted to keep, etc.

user2696128

Posted 2009-07-14T14:58:15.550

Reputation: 141

11

Reset the index and the head to origin/master, but do not reset the working tree:

git reset origin/master

user811773

Posted 2009-07-14T14:58:15.550

Reputation:

I personally found this to be most useful. It then keeps your working tree so you can check it in again. For my issue, I had the same files deleted as being added so it was stuck. Weird, I know. – Jason Sebring – 2014-01-04T21:03:59.443

11

I know of a much easier and less painful method:

$ git branch -m [branch_to_force_pull] tmp
$ git fetch
$ git checkout [branch_to_force_pull]
$ git branch -D tmp

That's it!

ddmytrenko

Posted 2009-07-14T14:58:15.550

Reputation: 638

9

Requirements:

  1. Track local changes so no-one here ever loses them.
  2. Make the local repository match the remote origin repository.

Solution:

  1. Stash the local changes.
  2. Fetch with a clean of files and directories ignoring .gitignore and hard reset to origin.

    git stash --include-untracked
    git fetch --all
    git clean -fdx
    git reset --hard origin/master
    

vezenkov

Posted 2009-07-14T14:58:15.550

Reputation: 2 242

9

I read through all the answers but I was looking for a single command to do this. Here is what I did. Added a git alias to .gitconfig

[alias]
      fp = "!f(){ git fetch ${1} ${2} && git reset --hard ${1}/${2};};f"

Run your command as

git fp origin master

equivalent to

git fetch origin master
git reset --hard origin/master

Venkat Kotra

Posted 2009-07-14T14:58:15.550

Reputation: 7 101

6

This is the best practice for reverting changes:

  • git commit Commit your staged changes so they will be saved in the reflog ( see below )
  • git fetch Fetch the latest upstream changes
  • git reset --hard origin/master Hard reset to the origin master branch

The reflog records branches and other references being updated in the local repository. Or simply put - the reflog is the history of your changes.

So it's always a great practice to commit. Commits are appended to the reflog which ensures you will always have a way to retrieve the deleted code.

Jordan Georgiev

Posted 2009-07-14T14:58:15.550

Reputation: 200

6

Don't use git reset --hard. That will wipe their changes which may well be completely undesirable. Instead:

git pull
git reset origin/master
git checkout <file1> <file2> ...

You can of course use git fetch instead of git pull since it clearly isn't going to merge, but if you usually pull it makes sense to continue to pull here.

So what happens here is that git pull updates your origin/master reference; git reset updates your local branch reference on to be the same as origin/master without updating any files, so your checked-out state is unchanged; then git checkout reverts files to your local branch index state as needed. In cases where exactly the same file has been added on live and on upstream master, the index already matches the file following the reset, so in the common case you don't need to do git checkout at all.

If the upstream branch also contains commits which you want to apply automatically, you can follow a subtle variation on the process:

git pull
git merge <commit before problem commit>
git reset <problem commit>
git checkout <file1> <file2> ...
git pull

Jim Driscoll

Posted 2009-07-14T14:58:15.550

Reputation: 700

5

I used this command to get rid of the local files preventing me from doing a pull/merge. But be careful! Run git merge … first to see whether there are only those files you really want to remove.

git merge origin/master 2>&1 >/dev/null | grep ^[[:space:]] | sed s/^[[:space:]]//g | xargs -L1 rm
  • git merge lists among other things all those files. They are prepended by some white-space.
  • 2>&1 >/dev/null redirects the error output to the standard one so it is picked up by grep.
  • grep ^[[:space:]] filters only the lines with file names.
  • sed s/^[[:space:]]//g trims the white-space from the beginning.
  • xargs -L1 rm calls rm on each of those files, deleting them.

Handle with care: Whatever git merge outputs, the rm will be called for every line beginning with a white-space.

Glutexo

Posted 2009-07-14T14:58:15.550

Reputation: 328

5

I was trying to use the Material2 branch on the Angular2-Webpack-Starter and had a heck of a time. This was the only way I could download and use that branch.

git clone --depth 1 https://github.com/angularclass/angular2-webpack-starter.git

cd angular2-webpack-starter/

git checkout -b material2

Open the project folder and delete all non-hidden files and folders. Leave all the hidden ones.

git add .

git commit -m "pokemon go"

git reset --hard

git pull origin material2

(When the editor pops up, hit ':wq', and then press Enter)

Now you are ready.

Helzgate

Posted 2009-07-14T14:58:15.550

Reputation: 4 758

3

You could ignore that file with a file in your project base folder:

.gitignore

public/images/*

Then pull the changes and then remove that line from your gitignore file.

Daniel Gaytán

Posted 2009-07-14T14:58:15.550

Reputation: 178

It contradicts Tierlieb's answer.

– Piotr Dobrogost – 2011-12-11T19:26:03.040

1

On Windows, do this single command:

git fetch --all & git reset --hard origin/master

Luca C.

Posted 2009-07-14T14:58:15.550

Reputation: 3 170

0

if you want to reset to the remote tracking branch in a generic way use:

git fetch
git reset --keep origin/$(git rev-parse --abbrev-ref HEAD)

if you want to reset your local changes too:

git fetch
git reset --hard origin/$(git rev-parse --abbrev-ref HEAD)

warch

Posted 2009-07-14T14:58:15.550

Reputation: 812

0

1: Reset to a previous commit

git reset --hard HEAD

2: Delete Untracked Files

git clean -f

3: Pull the commits

git pull

Sources:

abhijithvijayan

Posted 2009-07-14T14:58:15.550

Reputation: 27