Deleting large files from Git without breaking Jira and Crucible integrations

[Paul O'Flynn]

So I've done the research on removing these big files from history but it of course involves re-writing history and thus re-writing commit IDs. If I were to push this new modified repo to BitBucket it would presumably break the "Commits" links in (on-premise) JIRA and (on-premise) Crucible. I don't suppose anyone has tried this before and has any suggestions? This is not an open-source project. Thanks.

[Boris Berenberg]

For JIRA, all linking is done based on Issue Keys in Commit Messages. So everything should actually work just fine. I am not sure about Crucible. 

[Paul O'Flynn]

Are you sure about that? I know it looks at the commit message for the Jira ID but I assume it does that only as new commits come in and forever more after that it links the two in a more direct manner. If it was always only using the commit msg then Jira would have to scan every commit msg in the repo every time a Jira ticket was opened for viewing. Thanks for the reply BTW.

[Boris Berenberg]

What you will need to do in JIRA is disable smart commits, then go to manage the repos in your DVCS plugin page, then Shift+Click on the Refresh repo icon, and do a Full refresh. Then you can re-enable smart commits. But yeah, everything will just be re-indexed based on the commit messages. I spoke with the Crucible team, and they said they would expect all of the "new" commits to get picked up as new commits, not replaced on the old ones. Do you have https://confluence.atlassian.com/display/CRUCIBLE/Storing+all+revisions+under+review enabled?

[Paul O'Flynn]

Great, thx for the very informative answer. Regarding Crucible, if we turned this on would the new commit IDs (from old commits) _not_ be picked up and the review content still visible? TIA

[Boris Berenberg]

The old reviews should be orphaned basically with diffs in them but no commits. I would highly recommend testing this in a staging environment to be sure how it will work with your exact dataset and config.

[Paul O'Flynn]

Cool, thanks man.