It is rather interesting that Atlassian has taken the path in which they will simply delete people's repos. The data in the repo does not belong to them. Many of us are paying for private repos, every month, for every user and for years on end. Seems like a massive violation of customer trust. It is also possible that even those who host git repos here will need to rethink their usage of bitbucket as this shows that Atlassian may delete them too for any reason they choose.
I don't get it. So basically we need to create new projects due to change of the repository type. I do not want to be forced to setup everything from scratch, can you please add an option for one button conversion? I mean I can easily import from hg bitbucket to github, you yourself know there are plugins that could be used for that so it has to be possible
Because now I understand we have to:
1 rename my hg project
2 create a git project with my name
3 push the source to it once transformed to git
4 setup all docs, issues, wikis, webhooks, variables and everything else that there was previously hoping we didnt forget something
5 tryout whether pipelines work as they did and hope for the best
Would be a lot simpler to
1. Press a button to convert project to git project
So what happens to my issue trackers and wikis? Do I have to recreate hundreds of issues and wiki pages from scratch across multiple repositories or can I replace the existing hg repos with git repos without affecting the other components?
Removing the features that make your services unique is certainly an interesting business strategy. Good luck with that. Thanks for hosting my little projects for a while.
Question, since there is no option to convert a repo in-place: let's say I delete a repo, convert it to Git on my computer, can I re-recreate a Git repo *with the same name* and upload the Git stuff? Or is the name no longer available once you delete a repo?
I have used bitbucket purely for their mercurial support. Because they supported mercurial, I've been prepared to tolerate otherwise subpar service, like continuous integration to azure (solved via git intermediates, funnily enough), and I needed azure because of their flaky pipelines (outstanding tickets on this that are 4+ months old, still no solution and them asking me do the troubleshooting), plus flaky integrations (e.g. hello disappearing Trello boards). There are systemic issues in their offerings.
The upshot is this, if I go git, I'm going GitHub as they offer superior integrations. I'll be taking with me: my $$, my 40+ repos (some public open source, some private) and, most importantly, my recommendations that other dev's try Atlassian.
Atlassian absolutely has the right to make this call, but they're not the git option they think they are. Nor, I suspect, will it turn out the resources they recover (from dropping support for a VC system with 3% usage) allow them to solve their systemic defects, or be more competitive against the Microsoft (GitHub) behemoth.
The Github import of Mercurial -> Git is pretty good. You can import and even maintain your users by "attaching" the imported user names to Github user names. However, you need to get your users to make Github accounts first before you do the import. Also, it is probably best to temporarily mark your Bitbucket repo as private, import it as a private repo (as that allows you to attach users) and the mark the as public if needed. Overall, the process is easy but perhaps tedious. I do not know about wikis or issue trackers etc.
However, very clearly Atlassian does not care about users and even git users on Bitbucket should be worried about Atlassian deleting their private data one day. Best to move to some other provider.
I wanted to write something elaborate but I'm too frustrated right now. 7 years of data and the only reason of using BitBucket in the first place was Hg support.
We will not be doing business again any time in the future. Bye.
[I saved a copy of this post on a different cloud provider in case you decide to sunset support for English fonts in the future and retroactively kthxbye delete everything using one!]
Unless Atlassian make an easy way to migrate BitBucket hg repos to BitBucket git repos, without losing the issue tracker and pull request history, then I don't see the point staying on BitBucket. Add me to the list of users who chose BitBucket for mercurial support, and is now preparing to move everything to GitHub.
I'm also appalled at the decision to just delete everything. At the very least you could convert the mercurial repositories, issue trackers, and pull requests to static HTML pages, so that the history of repositories is preserved. The loss of old/unmaintained code is a real issue facing the software development community, and I expected Atlassian to be more socially responsible than just deleting everything.
It is shame. To delete customers repos what were paid just to make it "control system of choice for the industry" it is insulting customers, who were trusting to bitbucket for a long time. Everybody can be sure from now that once industry will choose another "control system" , bitbucket will delete GIT repos next time with no any doubts and respect to customers who were supporting them all whose years. Bitbucket does not care that some repositories were created long ago and have long history. They do not care that mercurial for some customers are more easier to use. They even did not note what is the percentage of users that have at least one mercurial project. Just they do not care. There are only two ways for me: 1. to move everything to local server. 2 . To waste time and convert mercurial projects to git and move to github.
Several teams of which I am or have been a member have used Bitbucket precisely because of its support for Mercurial. Over the years, we have been favorably impressed by Atlassian, so it was a real shock to learn that support for Mercurial is being dropped in a relatively short time-frame, without even the option of paying for Mercurial-based services. Perhaps Atlassian can rethink this possibility, even if there is an unwillingness to continue supporting Mercurial indefinitely. (Nine months is not a long time when there are multiple repositories and multiple dependencies.)
Given Atlassian's expertise in hosting Mercurial-based services, it seems to me that the marginal cost for Atlassian to continue hosting Mercurial repositories would be relatively small, so that the price to cover costs should be quite reasonable. Here's hoping Atlassian will think of this as a business opportunity as well as a way to regain the respect that is about to be lost if this bullying behavior is not reversed.
Thanks for nothing Atlassian. You were a real Mercurial temple in the beginning of the last decade. Now that you have accumulated a lot users you decide that it's better to get rid of Mercurial forcing all users to move away or force them to use Git. Git is not a standard, it's not because it has a big market share that all others SCMs must die. While not using Windows everywhere since it has the biggest market share? You've made your most stupid decision ever.
By the way, Mercurial has hundreds of contributions per week. From contributors and a large part from Mozilla and Facebook. I hope you're not thinking Mercurial is dead, at least.
I hope people will one day realize that hosting data on private clouds is a bloody stupid idea. Shame on you Atlassian.
Why not just to switch public Hg-repos into a read-only mode?
I understand that Atlassian has to spend some money to support an infrastructure for those read-only projects. But that keeps all references to Hg-repos, issues, Wiki-pages, PR-discussions, and tarballs valid. And there are plenty of such references on the Internet.
I can't imagine that all of them become invalid next year just because of a single decision of some Atlassian top-manager(s).
In view of the constant closures and restrictions of various services, I see a solution in self-hosted, because all services have shown their low reliability ((
I worked with students on bitbucket for many years. Most of the repositories are kept for history. Atlassian now wants to delete this information. Even if I convert the repositories to Git, I will lose the history of comments and issues. All links in my articles and documents like htpps://bitbucket.org/foo/bar lose their relevance.
Question: Can I then trust Atlassian with my information?
Atlassian, please provide a one click migration tool for converting the existing repositories from Mercurial to Git. Otherwise it would be just as interesting as moving to Github or Gitlab entirely...
I hope you folks understand that you're an expandable cost to Atlassian. Their core business revolves around Jira, not BB tickets. As they said, you're also only around 1% of their users, and they have little interest in keeping a vocal minority. They're going to prefer happy (or complacent) users. They kinda _want_ to drive you away, if you're going to cost them extra effort.
They wouldn't do this without discussing with PR and legal departments. It's just cheaper for them to piss you off and drive you away than to give you proper migration tools.
If you use Jira, or Pipelines, or a lot of BB integrations, then wait a few months, see what tooling they add, use it, or migrate manually. If you don't care about any of that ecosystem, then hooray, other Hg providers will welcome your business.
I hope you folks understand that you're an expandable cost to Atlassian. *snip* They kinda _want_ to drive you away, if you're going to cost them extra effort.
I fear this is very close to the truth. Let's face it, Mercurial users have been cut adrift with no provision for migration. The competition offer much smoother import facilities and are much larger.
Let's face it, how many users want the badges that we seem to be earning by making forum posts relative to doubling the available platforms for this useful service. Heck, they've time and money to be pushing surveys about coffee into my facebook feed - if they wanted Mercurial users they could easily keep them, but it's not working out for them on the bottom line.
It's a shame, they've bought up a useful service and now just want to run it as an add-on for their core products.
I've tried hg-fast-export and run into the #132 problem (https://github.com/frej/fast-export/issues/132), like @cayhorstmann I used the tags/v180317 solution listed there. After the author mapping process I got something to work but haven't tested the resulting gut repo to any great extent.
However we also make extensive use of subrepos (mainly hg subrepos, but some git ones too) and so would want to make use of some of the commits which support this that happened in January this year.
I'm a Java programmer and dont know much about python and pip so some of the other things listed in #132 are getting out of my comfort zone. #131 and #133 also make interesting reading here and add some context (I can see Frej's point of view given limited resources/manpower).
It dates back from 2016 and could do with an update(?), possibly from someone who understands the python world better than me.
But it got me thinking.... A docker image could be run as part of a bitbucket pipeline - a pipeline that runs hg-fast-export and it could also via the REST APIs have code to migrate issues, may be even pull requests? Other comments have asked for a "push button" converter and @AmberVH has responded that use cases were too varied and vast assumptions on workflows were needed. So how about giving us a (git) repo with bitbucket pipeline that we can fork and customize to our needs/workflows that would do a basic form of conversion as a starting point. Then users could fork, customize and share their conversion approaches using bitbucket?
I hope you folks understand that you're an expandable cost to Atlassian. *snip* They kinda _want_ to drive you away, if you're going to cost them extra effort.
I fear this is very close to the truth. Let's face it, Mercurial users have been cut adrift with no provision for migration. The competition offer much smoother import facilities and are much larger.
Let's face it, how many users want the badges that we seem to be earning by making forum posts relative to doubling the available platforms for this useful service. Heck, they've time and money to be pushing surveys about coffee into my facebook feed - if they wanted Mercurial users they could easily keep them, but it's not working out for them on the bottom line.
It's a shame, they've bought up a useful service and now just want to run it as an add-on for their core products.
Elevate your skills and enroll in the Compass learning path today! 💻
This learning path offers short, self-paced courses to help software teams systematically manage components and services. Compass provides a unified interface for tracking distributed software architecture and team collaboration.
531 comments