Git commit error: fatal: Repository has been updated, but unable to write new_index file.

Git commit error: fatal: Repository has been updated, but unable to write new_index file.

[Kirill Likhodedov]

Hello,

I'm developing a Git integration with an IDE. Several users of the program report the following error which was returned by Git during commit:

fatal: Repository has been updated, but unable to write
new_index file. Check that disk is not full or quota is
not exceeded, and then "git reset HEAD" to recover.

Here is the command which IDE executes to make a commit: 

# git commit --only -F <message file> -- <several files chosen to commit>

The interesting thing is that commit was actually successful, i.e. all files are committed, git status reports correctly, no other mistakes were noticed by users.
Here is the bugreport with some more details: http://youtrack.jetbrains.com/issue/IDEA-82108
One of the users says, that the situation happens quite often.

Users say that the thing has never happened with them while they were using Tortoise Git, so I'm pretty sure, that it is my bug caused my some misusage of Git (although I asked users to perform more experiments).

Unfortunately, I didn't find any information about the possible cause of this error.

Could you please help me to figure out why it can happen? What situation can lead to this error?

Thanks a lot.

----------------------------------
Kirill Likhodedov
JetBrains, Inc
http://www.jetbrains.com
"Develop with pleasure!"

Git commit error: fatal: Repository has been updated, but unable to write new_index file.

[Kirill Likhodedov]

Reposting, since there were no replies - probably, the message was missed.
I would appreciate help in understanding what the Git error could be cause by.
Thanks!

-------

Hello,

I'm developing a Git integration with an IDE. Several users of the program report the following error which was returned by Git during commit:

fatal: Repository has been updated, but unable to write
new_index file. Check that disk is not full or quota is
not exceeded, and then "git reset HEAD" to recover.

Here is the command which IDE executes to make a commit: 

# git commit --only -F <message file> -- <several files chosen to commit>

The interesting thing is that commit was actually successful, i.e. all files are committed, git status reports correctly, no other mistakes were noticed by users.
Here is the bugreport with some more details: http://youtrack.jetbrains.com/issue/IDEA-82108
One of the users says, that the situation happens quite often.

Users say that the thing has never happened with them while they were using Tortoise Git, so I'm pretty sure, that it is my bug caused my some misusage of Git (although I asked users to perform more experiments).

Unfortunately, I didn't find any information about the possible cause of this error.

Could you please help me to figure out why it can happen? What situation can lead to this error?

Thanks a lot.

----------------------------------
Kirill Likhodedov
JetBrains, Inc
http://www.jetbrains.com
"Develop with pleasure!"

Re: Git commit error: fatal: Repository has been updated, but unable to write new_index file.

[Kevin]

This might look like some kind of race condition where multiple commands
are executed at the same time, causing the lock file to be hold.

Could you check if something like that is happening somewhere?

On Tue, Mar 27, 2012 at 11:35:08AM +0400, Kirill Likhodedov wrote:
> 
> Reposting, since there were no replies - probably, the message was missed.
> I would appreciate help in understanding what the Git error could be cause by.
> Thanks!
> 
> -------
> 
> Hello,
> 
> I'm developing a Git integration with an IDE. Several users of the program report the following error which was returned by Git during commit:
> 
> fatal: Repository has been updated, but unable to write
> new_index file. Check that disk is not full or quota is
> not exceeded, and then "git reset HEAD" to recover.
> 
> Here is the command which IDE executes to make a commit: 
> 
> # git commit --only -F <message file> -- <several files chosen to commit>
> 
> The interesting thing is that commit was actually successful, i.e. all files are committed, git status reports correctly, no other mistakes were noticed by users.
> Here is the bugreport with some more details: http://youtrack.jetbrains.com/issue/IDEA-82108
> One of the users says, that the situation happens quite often.
> 
> Users say that the thing has never happened with them while they were using Tortoise Git, so I'm pretty sure, that it is my bug caused my some misusage of Git (although I asked users to perform more experiments).
> 
> Unfortunately, I didn't find any information about the possible cause of this error.
> 
> Could you please help me to figure out why it can happen? What situation can lead to this error?
> 
> Thanks a lot.
> 
> ----------------------------------
> Kirill Likhodedov
> JetBrains, Inc
> http://www.jetbrains.com
> "Develop with pleasure!"
> --
> To unsubscribe from this list: send the line "unsubscribe git" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Git commit error: fatal: Repository has been updated, but unable to write new_index file.

[Ben Tebulin]

> fatal: Repository has been updated, but unable to write new_index
> file. Check that disk is not full or quota is not exceeded, and then
> "git reset HEAD" to recover.
>
> Here is the command which IDE executes to make a commit:
>
> # git commit --only -F<message file>  -- <several files chosen to
> commit>

I can confirm this issue. I saw this error more than once. And I'm 
working with cygwin Git on the command line - but having an IDE (VS in 
this case) opened in parallel.

I don't think the issue is about two parallel Git operations in 
parallel. I only started one. But more that some file of the working 
copy is in access. And Git seems to be much more sensible about those 
than i.e. subversion. (No file was "locked")

Sorry - but I cannot tell anything more except that I've experienced 
this issue too. Maybe this is a  windows specific issue? As a windows 
user I had some of them...

Re: Git commit error: fatal: Repository has been updated, but unable to write new_index file.

[Kevin]

One hint that this is a windows issue is that all the reporters seem to
use msysgit.

(added msysgit mailinglist as cc)

On Tue, Mar 27, 2012 at 01:51:48PM +0200, Ben Tebulin wrote:
> >fatal: Repository has been updated, but unable to write new_index
> >file. Check that disk is not full or quota is not exceeded, and then
> >"git reset HEAD" to recover.
> >
> >Here is the command which IDE executes to make a commit:
> >
> ># git commit --only -F<message file>  -- <several files chosen to
> >commit>
> 
> I can confirm this issue. I saw this error more than once. And I'm
> working with cygwin Git on the command line - but having an IDE (VS
> in this case) opened in parallel.
> 
> I don't think the issue is about two parallel Git operations in
> parallel. I only started one. But more that some file of the working
> copy is in access. And Git seems to be much more sensible about
> those than i.e. subversion. (No file was "locked")
> 
> Sorry - but I cannot tell anything more except that I've experienced
> this issue too. Maybe this is a  windows specific issue? As a
> windows user I had some of them...
> 
> --
> To unsubscribe from this list: send the line "unsubscribe git" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Git commit error: fatal: Repository has been updated, but unable to write new_index file.

[Kirill Likhodedov]

27.03.2012, в 12:36, Kevin:

> This might look like some kind of race condition where multiple commands
> are executed at the same time, causing the lock file to be hold.
> 
> Could you check if something like that is happening somewhere?

I thought about it, but we've got a software lock that prevents such a collision, and it seems to work pretty good: we've never had lock problems in any other cases, and in this case the problem also seems to be quite rare - it is experienced only by a few of users, although rather frequently.
Moreover, there are no parallel Git activities in this case anyway, only git add followed by git commit.

Btw, lock-conflicting errors are usually reported by Git as 
fatal: Unable to create 'project_path/.git/index.lock': File exists.


> 
> On Tue, Mar 27, 2012 at 11:35:08AM +0400, Kirill Likhodedov wrote:
>> 
>> Reposting, since there were no replies - probably, the message was missed.
>> I would appreciate help in understanding what the Git error could be cause by.
>> Thanks!
>> 
>> -------
>> 
>> Hello,
>> 
>> I'm developing a Git integration with an IDE. Several users of the program report the following error which was returned by Git during commit:
>> 
>> fatal: Repository has been updated, but unable to write
>> new_index file. Check that disk is not full or quota is
>> not exceeded, and then "git reset HEAD" to recover.
>> 
>> Here is the command which IDE executes to make a commit: 
>> 
>> # git commit --only -F <message file> -- <several files chosen to commit>
>> 
>> The interesting thing is that commit was actually successful, i.e. all files are committed, git status reports correctly, no other mistakes were noticed by users.
>> Here is the bugreport with some more details: http://youtrack.jetbrains.com/issue/IDEA-82108
>> One of the users says, that the situation happens quite often.
>> 
>> Users say that the thing has never happened with them while they were using Tortoise Git, so I'm pretty sure, that it is my bug caused my some misusage of Git (although I asked users to perform more experiments).
>> 
>> Unfortunately, I didn't find any information about the possible cause of this error.
>> 
>> Could you please help me to figure out why it can happen? What situation can lead to this error?
>> 
>> Thanks a lot.
>> 
>> ----------------------------------
>> Kirill Likhodedov
>> JetBrains, Inc
>> http://www.jetbrains.com
>> "Develop with pleasure!"
>> --
>> To unsubscribe from this list: send the line "unsubscribe git" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Git commit error: fatal: Repository has been updated, but unable to write new_index file.

[Sylvain]

Hi,

I'm also experiencing this issue (msysGit 1.7.10-preview20120409 + TortoiseGit 
1.7.9.0 + Intellij IDEA 11).

Don't you think it could be related to parallel usage of msysGit + TortoiseGit?

I.e. after committing with msysGit (through IDEA or not), TortoiseGit would 
launch a background operation (such as updating index) that would mess with 
msysGit?

Just a guess, I don't know internal behaviour of msysGit/TortoiseGit.

Sylvain

Re: Git commit error: fatal: Repository has been updated, but unable to write new_index file.

[Heiko Voigt]

Hi,

On Tue, May 15, 2012 at 06:34:31AM +0000, Sylvain wrote:
> I'm also experiencing this issue (msysGit 1.7.10-preview20120409 + TortoiseGit 
> 1.7.9.0 + Intellij IDEA 11).
> 
> Don't you think it could be related to parallel usage of msysGit + TortoiseGit?
> 
> I.e. after committing with msysGit (through IDEA or not), TortoiseGit would 
> launch a background operation (such as updating index) that would mess with 
> msysGit?
> 
> Just a guess, I don't know internal behaviour of msysGit/TortoiseGit.

This is the wrong list for msysgit please see 

http://groups.google.com/group/msysgit

My guess is that some other program (your virus scanner or IDE ?) is
blocking writing on the index file. In msysgits git gui and the command
line we have a retry fallback that asks you whether you want to try
again. I do not know TortoiseGit I guess that they do not have such
thing and just fail.

Cheers Heiko