# Severe Regression in Git 1.21.0 -- not receiving any data

2 messages from 2019-05-31 to 2019-05-31. Participants: David, Beat Bolli.
Thread: https://gitlist.dev/t/51207

## David, 2019-05-31 11:31

Subject: Severe Regression in Git 1.21.0 -- not receiving any data
Message-ID: <04a82e31-2292-ef73-98c5-b9a5ff28e000@david-kremer.fr>
URL: https://gitlist.dev/e/04a82e31-2292-ef73-98c5-b9a5ff28e000%40david-kremer.fr

```
# Summary

I was trying to clone the AOSP source tree (Android Open Source Project) 
and I finally was able to formally identify what seems to be a severe 
regression (one that is blocking usage) in the last version of git (2.21.0).

# Steps to reproduce

The AOSP projects recommend to use ubuntu 14.04 as its build platform. 
The git version for this distro is :

ˋˋˋ
root@eb57f366840e:/android# git --version
git version 1.9.1
ˋˋˋ

In a docker image for this OS, I am able to execute:

ˋˋˋ
root@eb57f366840e:~# git clone 
https://android.googlesource.com/platform/manifest
Cloning into 'manifest'...
remote: Sending approximately 39.94 MiB ...
remote: Counting objects: 852, done
remote: Finding sources: 100% (27/27)
remote: Total 16017 (delta 4968), reused 16017 (delta 4968)
Receiving objects: 100% (16017/16017), 39.65 MiB | 9.04 MiB/s, done.
Resolving deltas: 100% (4968/4968), done.
Checking connectivity... done.
ˋˋˋ

Now, on my local Archlinux Box, the git version is 2.21.0 (latest).

The same command:

ˋˋˋ

git clone https://android.googlesource.com/platform/manifest
ˋˋˋ

Is launching a single thread that uses 100% of one CPU core FOREVER, 
until a timeout arises after like 40 minutes (the remote hung up 
unexpectedly). Someone on the #git channel (irc.freenode.net) was able 
to reproduce the exact same behaviour.

I must add that this of course arises only with the repo I mention in 
this email (github repositories are just fine). So probably a mismatch 
in protocol between git server/client ? Anyways, I'm far from having the 
skills to fix that thing.

I hope this report is useful, and in any case, I would like to know 
about the possible outcomes of this report.

I would be actually very surprised if someone did not notice that 
before, given that the latest git version is from february, but I can 
try anyway :)

Cheers,

David


```

## Beat Bolli, 2019-05-31 19:18

Subject: Re: Severe Regression in Git 1.21.0 -- not receiving any data
Message-ID: <a6a0e1be-8450-e61c-fe1c-911831004f66@drbeat.li>
URL: https://gitlist.dev/e/a6a0e1be-8450-e61c-fe1c-911831004f66%40drbeat.li
In-Reply-To: <04a82e31-2292-ef73-98c5-b9a5ff28e000@david-kremer.fr>

```
On 31.05.19 13:31, David wrote:
> # Summary
> 
> I was trying to clone the AOSP source tree (Android Open Source Project)
> and I finally was able to formally identify what seems to be a severe
> regression (one that is blocking usage) in the last version of git
> (2.21.0).
> 
> # Steps to reproduce
> 
> The AOSP projects recommend to use ubuntu 14.04 as its build platform.
> The git version for this distro is :
> 
> ˋˋˋ
> root@eb57f366840e:/android# git --version
> git version 1.9.1
> ˋˋˋ
> 
> In a docker image for this OS, I am able to execute:
> 
> ˋˋˋ
> root@eb57f366840e:~# git clone
> https://android.googlesource.com/platform/manifest
> Cloning into 'manifest'...
> remote: Sending approximately 39.94 MiB ...
> remote: Counting objects: 852, done
> remote: Finding sources: 100% (27/27)
> remote: Total 16017 (delta 4968), reused 16017 (delta 4968)
> Receiving objects: 100% (16017/16017), 39.65 MiB | 9.04 MiB/s, done.
> Resolving deltas: 100% (4968/4968), done.
> Checking connectivity... done.
> ˋˋˋ
> 
> Now, on my local Archlinux Box, the git version is 2.21.0 (latest).
> 
> The same command:
> 
> ˋˋˋ
> 
> git clone https://android.googlesource.com/platform/manifest
> ˋˋˋ
> 
> Is launching a single thread that uses 100% of one CPU core FOREVER,
> until a timeout arises after like 40 minutes (the remote hung up
> unexpectedly). Someone on the #git channel (irc.freenode.net) was able
> to reproduce the exact same behaviour.
> 
> I must add that this of course arises only with the repo I mention in
> this email (github repositories are just fine). So probably a mismatch
> in protocol between git server/client ? Anyways, I'm far from having the
> skills to fix that thing.
> 
> I hope this report is useful, and in any case, I would like to know
> about the possible outcomes of this report.
> 
> I would be actually very surprised if someone did not notice that
> before, given that the latest git version is from february, but I can
> try anyway :)
> 
> Cheers,
> 
> David

Can you try

    GIT_TRACE_PACKET=/tmp/packet.txt GIT_TRACE=/tmp/trace.txt git clone
https://android.googlesource.com/platform/manifest

?

The new files in /tmp help you see where the operation stalls.

Warning: the output will be large!


Beat

```