2

We have a code review process in place where a developer sends out a patch file with his changes to the team. After reviewing it, he is instructed to commit or make changes and resend.

How can we ensure that what has been "committed" is exactly what has been "approved" - i.e., if he makes subsequent changes without approval and commits those, how can I detect those?

I have the original 'patch file' at my end, but:

  • How can I 'generate' something similar between the two committed versions and
  • Is it viable to compare those two files?
Machavity
  • 30,841
  • 27
  • 92
  • 100
siliconpi
  • 8,105
  • 18
  • 69
  • 107

4 Answers4

1

The easiest way is that The reviewer commit the changes he approved. I've seen some old fashion team (a while ago) where only the project manager could commit to production repository. The thing is, as he has to commit too much stuff he didn't really check anything so that was a bit pointless.

mb14
  • 22,276
  • 7
  • 60
  • 102
  • @abatishchev - I find it an interesting idea, but I'm uncomfortable with it. The logs will show only one person committing and it wont really help when figuring out things later on (when things go wrong). – siliconpi Sep 22 '11 at 09:29
  • @sid I understand you concerns. Will that work if instead of commit , the committer create a new branch and only the reviewer can merge it into the new one. That's what we do with git (via pull request) and that works really well (as git keep the initial autor through merge). Maybe it's time to switch to git then ;-) – mb14 Sep 22 '11 at 09:48
1

The simplest solution would be to put such patches to a branch which mean the developer will check in the code on the branch. Than a reviewer can check the code and merge the code back to a particular integration line. This makes sure the code checked in is exactly what has been checked. Furthermore this approach has the advanctage the suggestion which have been made will be documented in the version control as well.

khmarbaise
  • 92,914
  • 28
  • 189
  • 235
0

Reviewer should be a committer as well to make sure he commits the same he reviewed and approved.

abatishchev
  • 98,240
  • 88
  • 296
  • 433
0

I am guessing you could do this by applying the patch on the old revision and then comparing the same against the latest version of code for difference.

You could use svn diff for this.

You probably wanna take a look at this answer: How to make svn diff produce file that patch would apply, when svn cp or svn mv was used?

Community
  • 1
  • 1
Baz1nga
  • 15,485
  • 3
  • 35
  • 61
  • Hi zzzz - I think svn diff might work for me. I'm able to use: svn diff --revision NNN:MMM" but its displaying the files out of order (as compared to the actual patch)... Anyway, its a start! Thanks! – siliconpi Sep 22 '11 at 10:36