Git 2.19.1 crashes while git gc in Windows

[guoxiao]

• I was not able to find an open or closed issue matching what I'm seeing

I found previous issues like this which was fixed in 2.19.1, but git still crashes on my machine after I upgraded to the latest version.

Setup

• Which version of Git for Windows are you using? Is it 32-bit or 64-bit?

$ git --version --build-options

git version 2.19.1.windows.1
cpu: x86_64
built from commit: 11a3092e18f2201acd53e45aaa006f1601b6c02a
sizeof-long: 4
sizeof-size_t: 8

• Which version of Windows are you running? Vista, 7, 8, 10? Is it 32-bit or 64-bit?

$ cmd.exe /c ver

Microsoft Windows [Version 10.0.17134.320]

• What options did you set as part of the installation? Or did you choose the
  defaults?

# One of the following:
> type "C:\Program Files\Git\etc\install-options.txt"
> type "C:\Program Files (x86)\Git\etc\install-options.txt"
> type "%USERPROFILE%\AppData\Local\Programs\Git\etc\install-options.txt"
$ cat /etc/install-options.txt

Editor Option: VIM
Custom Editor Path:
Path Option: Cmd
SSH Option: OpenSSH
CURL Option: WinSSL
CRLF Option: CRLFCommitAsIs
Bash Terminal Option: MinTTY
Performance Tweaks FSCache: Enabled
Use Credential Manager: Enabled
Enable Symlinks: Enabled
Enable Builtin Rebase: Enabled
Enable Builtin Stash: Enabled

• Any other interesting things about your environment that might be related
  to the issue you're seeing?

I use powershell with poshgit plugin.

Details

• Which terminal/shell are you running Git from? e.g Bash/CMD/PowerShell/other

Powershell

• What commands did you run to trigger this issue? If you can provide a
  Minimal, Complete, and Verifiable example
  this will help us understand the issue.

git gc

• What did you expect to occur after running these commands?

It works without error.

• What actually happened instead?

It crashed with message:

Enumerating objects: 952556, done.
error: failed to run repack03/952556)

• If the problem was occurring with a specific repository, can you provide the
  URL to that repository to help us with testing?

It occurs on a internal repo that I cannot share here. The crash dump file is here:
git.exe.14828.zip

I tried the latest snapshot at https://wingit.blob.core.windows.net/files/index.html but the error still exists.

[orgads]

Happened to me too.

[dscho]

This looks quite similar to #1839 with the notable difference that this here ticket does not provide a Minimal, Complete & Verifiable Example. Could somebody provide one?

[dscho]

@orgads could you please come up with an MCVE? If there is no way to reproduce this reliably, we cannot fix this.

[orgads]

Didn't have this so far with 2.19.2.

[dscho]

843931b was already integrated in v2.19.1. I do not see anything relevant to gc that is different between v2.19.1 and v2.19.2...

And the symptom really looks like the one of the bug fixed in said commit. So I am really at a loss here.

[orgads]

Good point. I have no idea then. If it happens again, I'll try to investigate further.

[dscho]

On Thu, 29 Nov 2018, Orgad Shaneh wrote: Good point. I have no idea then. If it happens again, I'll try to investigate further.

👍 Thank you!

[Neumann-A]

seems the issue is still there:

C:\Sources\Repos\SVN\mpiThirdParty>git --version
git version 2.20.0.windows.1

C:\Sources\Repos\SVN\mpiThirdParty>git gc
Enumerating objects: 49480, done.
fatal: failed to run repack4/49480)

C:\Sources\Repos\SVN\mpiThirdParty>ver

Microsoft Windows [Version 6.1.7601]

If you tell me how to build git by hand using VS Code or Visual Studio (or CMake). I could try debugging it or finding out where exactly it crashes.
figured it out. Just checkout vs/master ;)

[Neumann-A]

Ausnahmefehler bei 0x0000000077122B04 (ntdll.dll) in git.exe: 0xC0000005: Zugriffsverletzung beim Schreiben an Position 0x0000000000000024
(Access violation exception for those who don't speak german)

If I pause git in the debugger i see the following in the main thread:

other thread(s) just seems to wait.

I cannot see anything obviously wrong with it, so i need a little bit more help with debugging it.

[ikmplx]

I can reproduce crash with following steps:

# Initialize repo
git init

# Generate 3.6 GB of pseudorandom data
cat /dev/zero | openssl enc -aes-128-ctr -pass pass:"qwerty" -nosalt | dd bs=1M count=3600 iflag=fullblock of=test

git add test
git repack

gdb stacktrace:

(gdb) run
Starting program: D:\git-sdk-64\usr\src\git\git.exe "--exec-path=D:/git-sdk-64/usr/src/git" pack-objects --delta-base-offset .git/objects/pack/.tmp-13040-pack --keep-true-parents --honor-pack-keep --non-empty --all --reflog --indexed-objects --unpacked --incremental
[New Thread 5004.0x23bc]
[New Thread 5004.0x2838]
[New Thread 5004.0x2124]
[New Thread 5004.0x33b4]
14:48:36.163521 exec-cmd.c:236          trace: resolved executable dir: D:/git-sdk-64/usr/src/git
14:48:36.167013 git.c:418               trace: built-in: git pack-objects --delta-base-offset .git/objects/pack/.tmp-13040-pack --keep-true-parents --honor-pack-keep --non-empty --all --reflog --indexed-objects --unpacked --incremental
[New Thread 5004.0x97c]

Enumerating objects: 1, done.
[Thread 5004.0x97c exited with code 0]
[New Thread 5004.0x16e4]

Thread 1 received signal SIGSEGV, Segmentation fault.
0x00007ffbb371d979 in ntdll!TpSetWait () from C:\WINDOWS\SYSTEM32\ntdll.dll
(gdb)
(gdb) bt
#0  0x00007ffbb371d979 in ntdll!TpSetWait ()
   from C:\WINDOWS\SYSTEM32\ntdll.dll
#1  0x00007ffbb37183e7 in ntdll!RtlEnterCriticalSection ()
   from C:\WINDOWS\SYSTEM32\ntdll.dll
#2  0x00007ffbb3718300 in ntdll!RtlEnterCriticalSection ()
   from C:\WINDOWS\SYSTEM32\ntdll.dll
#3  0x000000000047026c in oe_get_size_slow (pack=0x73fc20 <to_pack>,
    e=0x3684248) at builtin/pack-objects.c:1996
#4  0x000000000046ba17 in oe_set_size (pack=0x73fc20 <to_pack>, e=0x3684248,
    size=3774873600) at ./pack-objects.h:362
#5  0x000000000046f957 in check_object (entry=0x3684248)
    at builtin/pack-objects.c:1655
#6  0x000000000046ff53 in get_object_details () at builtin/pack-objects.c:1873
#7  0x0000000000471c01 in prepare_pack (window=10, depth=50)
    at builtin/pack-objects.c:2615
#8  0x00000000004746f5 in cmd_pack_objects (argc=0, argv=0x3670a08,
    prefix=0x0) at builtin/pack-objects.c:3491
#9  0x0000000000402e3b in run_builtin (p=0x680700 <commands+1728>, argc=11,
    argv=0x3670a08) at git.c:421
#10 0x00000000004031f6 in handle_builtin (argc=11, argv=0x3670a08)
    at git.c:653
#11 0x0000000000403421 in run_argv (argcp=0x136fdc0, argv=0x136fd68)
    at git.c:707
#12 0x000000000040382f in cmd_main (argc=11, argv=0x3670a08) at git.c:829
#13 0x00000000004bfa6f in main (argc=13, argv=0x36709f8) at common-main.c:45

$ git --version --build-options
git version 2.20.0.windows.1
cpu: x86_64
built from commit: 95155834166f64fe9666f2c0a4909f076080893a
sizeof-long: 4
sizeof-size_t: 8

[dscho]

Good news: this looks like it might be fixed by https://public-inbox.org/git/nycvar.QRO.7.76.6.1901181150450.41@tvgsbejvaqbjf.bet/T/#t.

[peteri]

I've just run into the same issue, applying the patch mentioned by @dscho did fix the problem (although the linked thread suggests the patch needs a bit more work).
NOTE if you apply the same patch locally you'll need to make sure the patched version of git is in your path as the "git gc" respawns git (to run in the Visual Studio debugger I patched exec-cmd.c to hard code the path back to my git.exe so I could test)

[dscho]

@peteri thanks for testing. The patch advanced already to git.git's master and will therefore be part of v2.21.0. According to https://tinyurl.com/gitCal, Git v2.21.0-rc0 should be out today, let's see how fast I can publish a Git for Windows v2.21.0-rc0 after that.

[peteri]

Just as an FYI 2.21.0 has fixed the particular GC problem I was seeing.