Re: Bug in lineendings handling that prevents resetting checking out, rebasing etc

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Thanks for looking to it - git attributes was added in
4b0863a8b834c5804fd3a568ed56ff85b27acdeb

The file in question was added in 17ca75f0a8c25f321f2b63bc3b9c065ff91adc23

So you mean to say that because a gitattributes was added after the
fact this resulted in an illegal state?

But _shouldn't git reset --hard work anyway?_ That's the buggy part.

As for fixing it - not sure what is the best course of action here.
probably issuing `git add --renormalize .` and committing that to the
stable (dev) branch will fix this for future checkouts/rebases but
IIUC won't do nothing for older commits - so checking out a commit
before the fix one, ghit will see this file as changed and then
completely refuse to go back to another branch

This seems a bug - as illegal as the state of the file is, shouldn't
git reset always work?

Thanks! (will be out for a bit but really looking forward to your replies)
On Fri, Dec 14, 2018 at 4:32 PM Torsten Bögershausen <tboegi@xxxxxx> wrote:
>
> On Fri, Dec 14, 2018 at 04:04:15PM -0500, Mr&Mrs D wrote:
> > Hi all,
> >
> > I maintain a python project you can clone from:
> >
> > git@xxxxxxxxxx:wrye-bash/wrye-bash.git
> >
> > For reasons unknown git sees a particular file as changed
> > (Mopy/Docs/Bash Readme Template.html, sometimes others too). This file
> > was probably committed to the svn repository this git repo was created
> > from with CRLF line endings. When we moved to git we added a
> > gitattributes file (
> > https://github.com/wrye-bash/wrye-bash/blob/dev/.gitattributes ) and
> > this file was edited to explicitly state htms are text - all to no
> > avail. From time to time - on windows - as in when checking out an old
> > commit - git would see that file as changed. The only workaround that
> > worked for me was
> >
> >     git rm -r . --cached -q && git reset --hard
> >
> > For more details and discussion see this SO question I posted almost
> > five years ago:
> >
> > https://stackoverflow.com/questions/21122094/git-line-endings-cant-stash-reset-and-now-cant-rebase-over-spurious-line-en
> >
> > I used to work in windows and the bug was tolerable as there was that
> > workaround. Now I moved to mac and no workaround works anymore - we
> > have a special page on our wiki  with workarounds for this one btw:
> >
> > https://github.com/wrye-bash/wrye-bash/wiki/%5Bgit%5D-Issues-with-line-endings-preventing-checking,-merge,-etc
> >
> > Well after 5 years and countless hours trying to solve this I reach
> > out to you guys and girls - _this is a full-time bug in git line
> > endings handling_. When someone issues a git reset --hard this should
> > work no matter what - well it does not. So this bug may be really a
> > can of worms.
> >
> > Please someone clone this repo on linux or mac - probably just cloning
> > will have the files appear as changed (by the way hitting refresh on
> > git gui I have different sets of files appear as changed). If not then
> >
> > git checkout utumno-wip
> Thet commit is -excuse me if that sounds too harsh- is messed up.
> git status says
> modified:   Mopy/Docs/Bash Readme Template.html
>
> And if I dig into the EOL stuff, I run
> git ls-files --eol | grep  Readme | less
>
> And find a contradiction here:
> i/crlf  w/crlf  attr/text               Mopy/Docs/Bash Readme Template.html
>
> The attributes say "text" and the file has CRLF "in the repo",
> (techically speaking in the index) and that is an "illegal" condition
> in the repo, and not a bug in Git.
> I didn't try the rebase as such, sice the first problem needs
> to be fixed, before we try to move on.
>
> So, the old commits are problematic/illegal and they are as they are.
> Such a commit can not be fixed, whenever somebody checks it out,
> there will be a problem (or two, or none, depending on the timing,
> the file system...)
>
> We can not fix commits like b1acc012878c9fdd8b4ad610ce7eae0dcbcbcab0.
> We can make new commits, and fix them.
>
> We can fix one branch, and other branches, and merge them together.
> But rebase seems to be problamatic, at least to me.
> What exactly do you want to do?
>
> Can we agree to do a merge of 2 branches?
> Then I can possibly help you out.
>
>
>
>
>
> > git rebase -i dev
> >
> > and then select a commit to edit should be enough to trigger this bug
> >
> > Needless to say I am  well aware of things like `git add --renormalize
> > .` - but renormalizing is not the issue. The issue is that _files show
> > as changed and even a git reset --hard won't convince git that
> > nothing's changed_.
> >
> > $ git reset --hard
> > HEAD is now at e5c16790 Wip proper handling of ini tweaks encoding - TODOs:
> > $ git status
> > interactive rebase in progress; onto 02ae6f26
> > Last commands done (4 commands done):
> >    pick 3a39a0c0 Monkey patch for undecodable inis:
> >    pick e5c16790 Wip proper handling of ini tweaks encoding - TODOs:
> >   (see more in file .git/rebase-merge/done)
> > Next commands to do (19 remaining commands):
> >    edit a3a7b237 Amend last commit and linefixes:  ΕΕΕΕ
> >    edit 432fd314 fFF handle empty or malformed inis
> >   (use "git rebase --edit-todo" to view and edit)
> > You are currently editing a commit while rebasing branch 'utumno-wip'
> > on '02ae6f26'.
> >   (use "git commit --amend" to amend the current commit)
> >   (use "git rebase --continue" once you are satisfied with your changes)
> >
> > Changes not staged for commit:
> >   (use "git add <file>..." to update what will be committed)
> >   (use "git checkout -- <file>..." to discard changes in working directory)
> >
> > modified:   Mopy/Docs/Bash Readme Template.html
> >
> > Untracked files:
> >   (use "git add <file>..." to include in what will be committed)
> >
> > .DS_Store
> > .idea.7z
> >
> > no changes added to commit (use "git add" and/or "git commit -a")
> > $
> >
> > I really hope someone here can debug this
> > Thanks!




[Index of Archives]     [Linux Kernel Development]     [Gcc Help]     [IETF Annouce]     [DCCP]     [Netdev]     [Networking]     [Security]     [V4L]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Fedora Users]

  Powered by Linux