Closed
Description
- Gitea version (or commit ref): 1.1.0+635-gefa482d (built version of Rewrite migrations to not depend on future code changes #2604)
- Git version: 2.1.4
- Operating system: Raspian (Raspberry Pi Zero W)
- Database
- PostgreSQL
- MySQL
- MSSQL
- SQLite
- Can you reproduce the bug at https://try.gitea.io:
- Yes (provide example URL)
- No
- Not relevant
- Log gist:
Description
- I have a low amount of ram (512mb)
- Memory obtained stays very high and causes nothing to work
- Current memory usage stays very high (even though GC already ran)
- You need to view this commit (it's 36mb diff, so it's not shown): https://try.gitea.io/mrexodia/DarkSouls3.TextViewer/commit/629cf9b3d6b295bbcddf76d1f6167259b764d9dc
- Line where it happens:
Line 318 in 92123fe
Log
2017/10/06 21:18:00 [I] Log Mode: File(Info)
2017/10/06 21:18:01 [I] XORM Log Mode: File(Info)
2017/10/06 21:18:01 [I] Cache Service Enabled
2017/10/06 21:18:01 [I] Session Service Enabled
2017/10/06 21:18:01 [I] Mail Service Enabled
2017/10/06 21:18:01 [I] Register Mail Service Enabled
2017/10/06 21:18:01 [I] Notify Mail Service Enabled
2017/10/06 21:18:01 [I] Git Version: 2.1.4
2017/10/06 21:18:01 [I] SQLite3 Supported
2017/10/06 21:18:01 [I] Run Mode: Production
2017/10/06 21:18:04 [I] Listen: http://0.0.0.0:3000
2017/10/06 21:35:22 [...ules/context/repo.go:318 func1()] [E] GetTags: fork/exec /usr/bin/git: cannot allocate memory
2017/10/06 21:35:26 [...ules/context/repo.go:318 func1()] [E] GetTags: fork/exec /usr/bin/git: cannot allocate memory
2017/10/06 21:36:37 [...ules/context/repo.go:318 func1()] [E] GetTags: fork/exec /usr/bin/git: cannot allocate memory
2017/10/06 21:36:38 [...ules/context/repo.go:318 func1()] [E] GetTags: fork/exec /usr/bin/git: cannot allocate memory
2017/10/06 21:36:39 [...ules/context/repo.go:318 func1()] [E] GetTags: fork/exec /usr/bin/git: cannot allocate memory
Screenshots
Here is a video showing exactly how I reproduced the problem on a just-started gitea instance. I tried this about 4 times and it happens every time. It might be possible to reproduce on a system with more memory by cloning this repo multiple times and opening the offending diff a bunch of times as well, but I didn't test this yet.