Error when cloning: Smudge error

[bkodre]

• Gitea version (or commit ref):1.8.3
• Git version:2.22.0.windows.1
• Operating system:Windows server 2012 R2
• Database (use [x]):
  • PostgreSQL
  • MySQL
  • MSSQL
  • SQLite
• Can you reproduce the bug at https://try.gitea.io:
  • Yes (provide example URL)
  • No
  • Not relevant
• Log gist:

Description

When cloning repository i get error on downloading file tracked in lfs. Clone does not fail every time, but when it does it has always different file as a reason.

Error downloading object: Flash/lib/greensock.swc (1b0c435): Smudge error: Error downloading Flash/lib/greensock.swc (1b0c43569cad83be13b26dc11253aa696c42a8797d57a95983b3fde97bdbcf5d): batch response: Repository or object not found: http://hostname:3000/company/product.git/info/lfs/objects/batch
Check that it exists and that you have proper access to it

[lunny]

It seems v1.9 fixed that.

[bkodre]

When version 1.9 will be released.
I get this error verfy frequently

[zeripath]

Basically the error is saying that LFS was unable to download the file.

I think there are a number of possible reasons for this. However from your response it's not easy to determine which one of these is definitely responsible.

The most scary for you should be the situation that these LFS files are not in your repository. I think it's worth ensuring that they are and if not get someone who does have them to do a git lfs push --all (see here
git-lfs/git-lfs#1113 (comment) )

I have a pr open that will allow you to look at your repo and try to find all of the possible LFS pointers your repo has and compare them with the LFS files it knows about and those you can access allowing you to match these up and fix errors.

It's not necessarily Gitea's fault that this has happened if it has but converse of #732 (i.e. if you could merge LFS adding prs from forks) might have caused it. On 1.9 that should no longer happen but I seriously apologize if you were affected by that - all I can say is that getting LFS right is not as easy as it first appears.

[zeripath]

A good test might be to try git lfs fetch --all to see if there are LFS objects missing - check the Gitea logs as well on errors to see if

[bkodre]

i executed git lfs fetch --all command and it completed successfully ( Downloading LFS objects: 100% (1277/1277)).

The strange thing is that on my pc most of the times clone is done successfully, but my colleague gets smudge error almost every time.

[zeripath]

It might be worth them trying the fetch all technique - it could be that the repository still doesn't have all the objects but you have the missing ones.

Are you cloning over SSH or HTTP? It would be helpful to see the logs when these failures happen.

[bkodre]

HTTP was used for cloning.

He tried fetch all and got this error in lsf\logs(file attached):

batch response: Repository or object not found: http://hostname:3000/company/product.git/info/lfs/objects/batch
Check that it exists and that you have proper access to it
batch response: Fatal error: Server error: http://hostname:3000/company/product.git/info/lfs/objects/batch

20190711T102700.309593.log

[zeripath]

Ok that implies you have some LFS files missing from your repo - the Gitea logs would be more helpful however.

I don't really know how this has happened. I suspect the underlying issue #732 is to blame.

You need to from your computer, that has all the objects, do a git lfs push --all and then see if git lfs fetch --all on their machine works.

[bkodre]

I'm not sure that files are missing, because on my pc fresh clone most of times works ok.

I attached gitea logs:
gitea.log

I tried git lfs push --all and it failed:

Uploading LFS objects:  38% (3200/8409), 121 MB \| 0 B/s, done                    
batch response: Repository or object not found: http://hostname:3000/company/product.git/info/lfs/objects/batch
Check that it exists and that you have proper access to it

[zeripath]

Thanks for the Gitea log it looks like this could be a mysql connector issue! Someone else recently reported something similar.

[...s/context/context.go:238 func1()] [E] GetAccessTokenBySha: dial tcp 127.0.0.1:3307: connectex: Only one usage of each socket address (protocol/network address/port) is normally permitted.
2019/07/11 08:10:48 [...s/context/context.go:238 func1()] [E] UserSignIn: dial tcp 127.0.0.1:3307: connectex: Only one usage of each socket address (protocol/network address/port) is normally permitted.

[bkodre]

Any idea what can be done to resolve this issue with mysql connector?

[zeripath]

Ah it was you who reported it...

[bkodre]

Yes it was me, but i didn't found solution to it and i was not sure if these errors are connected.

[zeripath]

If you look at your log you get some authentication errors in the middle of these errors - I wonder if the db connection failure is causing authentication to fail.

[bkodre]

My credentials are saved in git client, so probably db connection is to blame for authentication fail.

[zeripath]

https://blogs.msdn.microsoft.com/dgorti/2005/09/18/only-one-usage-of-each-socket-address-protocolnetwork-addressport-is-normally-permitted/

[zeripath]

But we're not the first people to come across this issue and I wonder if following the techniques here might fix this: restic/restic#791

[bkodre]

I have set MaxUserPort to 65535 and TcpTimedWaitDelay to 30. Now it's working ok.

[zeripath]

Awesome.

[bkodre]

Thank you very much