Dont build commits with the same sha more than once#223
Dont build commits with the same sha more than once#223
Conversation
|
If no one opposes I'm going to deploy it tomorrow and see how it goes. |
|
Have you tested everything on staging? |
|
Can you also look into the jruby tests not running? |
No, I will be doing it tomorrow and if everything works ok I will deploy on production.
Sure, I will surely do it before deploying |
|
Once you deploy to Staging let me know, I would love to have a play around :) On 26/06/2013, at 2:14 PM, Piotr Sarnacki notifications@github.com wrote:
|
|
Sure, I guess I will do it today. On Wed, Jun 26, 2013 at 3:15 PM, Josh Kalderimis
Piotr Sarnacki |
The same commit may be available on 2 different branches. Currently we treat it as 2 separate commits, which is not optimal, build will be fired for each branch causing 2 identical builds. This is the first step to fix this problem.
When commit, which already exists on Travis is pushed, we will now add a new branch to branches list for an existing commit and an existing build.
This change is needed to allow keeping SQL functions in schema, but using structure.sql is good in general if you use branching - schema.rb keeps only last migration number, while structure.sql is just a schema dump, so it will save entire schema_migrations table.
Conflicts: .travis.yml
|
Sorry for leaving no further comments on this PR. After testing it on staging it turned out that there are problems with plugin which I used for using pg arrays and our jruby apps (travis-hub). I thought that the easiest fix would be to update to rails 4, which has a built-in support for pg arrays, but it turned out to be harder than I thought - although we only use AR, we have a few things, which are problematic. Unfortunately there were other pressing issues at that point and I left this unfinished. I rebased it to master and I will work on it later this week, especially because of #270 problems. Things to do:
|
|
This is outdated, so I'm closing it. I will use some parts of this (particularly API changes), but in general most of this PR will be addressed by #320 |
Sorry for so much of a delay with this, I was doing quite a lot of back & forth with the implementation.
I ended up using postgresql string arrays for keeping branch list for a commit and a build. This by far the best answer in my opinion, because it's dead simple and quite powerful and it does not need getting back to JOINs for getting info on last build on branch etc.
I will also push a small change in the API to start pushing
branchesas an array (but I will leavebranchas an alternative for backwards compatibility).