Git Off My Lawn

Git, the distributed source control system, is becoming very popular in the wider software world. Its proponents claim it revolutionises the way we make and test software, but is it suitable for cross-discipline game development?

For the creation of the App Store hit ‘Sonic Dash’, SEGA’s internal studio Hardlight experimented with using Git as their primary method of version control. Hear from the project lead and lead programmer about their practical experiences switching from Perforce, what went wrong, what went right, and how the team’s workflow changed.

Presented at Develop 2013 on July 11th 2013

Transcript

1. (JU 0GG
   .Z
   -BXO Pushing a team to the cutting edge
2. @lee_winder
3. @tenpn
4. PWFSWJFX reasons for the change the workflow we adopted the tools we used what problems we faced
5. TVSWFZ we carried out a to gathers people experience
6. TUBSUFE where we
10. we stored everything in Code Assets Documents SDKs
11. everyone was familiar with 11 out of 13 respondents said they were familiar or very familiar with p4
12. ... and everyone else uses
13. 6 months to release long update cycle planned 14+ Developers
14. programmers artists designers QA UFBN
    TUSVDUVSF
    
15. XIZ change things?
16. reasons we decided to use git
17.  JTPMBUFE feature development
18. XPSLJOH from home 
19. 6OJUZ and Perforce don’t mix 
20. DVSJPTJUZ  professional
21. price was never really a reason A large number of people thought it was though
22. (JU small and fast easy branching writable files
23. XPSLqPX Feature Branch QA Develop Release
25. http://tinyurl.com/gitflow
26. 2" Feature Branch Develop have veto
27. builds all branches merge manually builds develop daily
28. JU
    XBT
    B
    NJTUBLF
    UP DPOUJOVPVTMZ
    JOUFHSBUF BMM
    CSBODIFT
29. NBOBHFE branches should have been rather than a free-for-all
30. UPPMT are where you’ll sink or swim
31. in general the tools are QPPS on Windows
32. GSFF with client tools can get you a long way
33. TFSWFS but it’s worth spending money on your Gitolite
34. BSUJTUT
    XPSSJFE
    BCPVU UIF
    MBDL
    PG
    BO
    FYDMVTJWF
    DIFDLPVU but that’s not the fault of the tools really... A SCS with no exclusive checkout baffles me. I thought that was the point?
35. MBSHF assets? but what about
36. PSDs Audio Excel files Meshs automated conversion
37. GVUVSF
    QSPKFDUT
    XJMM VTF
    TNBMMFS
    TJCMJOH HJU
    SFQPTJUPSJFT but Perforce will always be required for some assets
38. TNPPUIMZ it would be boring if it all went
39. initial presentation covered the main differences
40. WPDBCVMBSZ knocked everyone sideways
41. submit get latest pull checkout stage commit push commit
42. DIFDL
    PVU
43. SFCBTF we completely ignored
44. NBOZ
    UP
    POF maybe SVN could have been easier to use as the 1 server - N clients logic is easier to understand than the "each client is also a server" logic I do not like the two stage process commit/push - I find it redundant and confusing we clearly didn’t explain the ‘commit to push’ workflow
45. I still don't understand how branches flow in the SourceTree view and that scares me.
46. QFSTPOBM
    CSBODIFT a number of developers kept insisting on
47. TDFOF
    NBOBHFNFOU XBT
    PVS
    CJHHFTU
    QSPCMFN exclusive checkout was not the answer
48. DPOTJTUFOU
    TUPSZ we should have stuck to a when speaking to the team
49. IPX
    UP
    JNQSPWF discipline evangelists uniform tool usage regular training consistent story
50. HPPE CBE the and the
51. HPPE consistently stable builds strong QA process flexible working styles remote working friendly
52. CBE didn’t provide enough training initially used poor tools unstructured branch creation server support was lacking
53. UP
    SFDBQ the experience was generally positive as the tools improve so will the workflow we’ll use git on all future projects we’ll spend more time educating people
54. 5IBOLT Lee Winder @lee_winder Andy Fray @tenpn