It's 2020 And GCC Has Finally Converted From SVN To Git

[alcalde]

Once things settle, it will be time for the next discussion.

Great; I'm working on my draft proposal right now that the project transition to Mercurial.

[duby229]

What exactly is "improper" about GitHub? It's exactly these self-hosted amateurish services that cause all kinds of troubles.

Which provider is bound to have more experience with running the services? The one that runs a couple dozen repos, or the one serving a couple million repos? I mean it's completely out of question, it's pointless to even argue about.

Not to mention various synergies resulting from a GitHub host, like people don't have to sign up a separate account just for GCC, they're already familiar with the interface etc. And I'm saying this as someone who runs GitLab for the company. GitLab is great for small to midsize teams, but if you want something serious, it's a whole lot of $$$. I also don't much like the interface. I mean it's decent, just not as good.

What's improper about GitHub is that it's practically a data miner for a specific large corporation with anti-trust and EEE aspirations and with a long varied history of accomplishing such....

[CommunityMember]

... but making the self-hosted instance the primary mirror is just foolish in this day and age.

I think as a first initial step (having full control) it *might* have made some (a little, maybe) sense (although anyone with infrastructure clue would have warned them against it precisely because the first thing everyone (and their brother) would do is clone this huge repo, resulting in lots of server side work as it compresses this repo again, and again, and again. After all the entire family has finished the initial clones, keeping things synced will likely take far less server side resources). But they really should strongly reconsider self-hosting unless they expect to move the service to their herd OS based server soon (in which case they may need to continue self hosting).

[anarki2]

Doesn't a extremely important project such as GCC have a proper server?

What about their own GitLab instance too?

What exactly is "improper" about GitHub? It's exactly these self-hosted amateurish services that cause all kinds of troubles.

Which provider is bound to have more experience with running the services? The one that runs a couple dozen repos, or the one serving a couple million repos? I mean it's completely out of question, it's pointless to even argue about.

Not to mention various synergies resulting from a GitHub host, like people don't have to sign up a separate account just for GCC, they're already familiar with the interface etc. And I'm saying this as someone who runs GitLab for the company. GitLab is great for small to midsize teams, but if you want something serious, it's a whole lot of $$$. I also don't much like the interface. I mean it's decent, just not as good.

[anarki2]

http://gcc.gnu.org/git/?p=gcc.git;a=summary

The GitHub mirror has successfully updated: https://github.com/gcc-mirror/gcc (that has everything, including the not-fetched by default refs, plus GitHub-specific refs/pull/* refs for pull requests people created on GitHub in the past). The GitHub mirror should keep automatically updating in future.

Please use it instead of the primary mirror for cloning since gcc.gnu.org can barely handle many people doing an initial checkout.

Or how about don't even bother with a gnu mirror? What's the point? We're way past the days when every single project had to host their own stuff. It just doesn't make much sense for either party anymore.

If GNU wants to have backups, that makes perfect sense, but making the self-hosted instance the primary mirror is just foolish in this day and age. GitHub gives them all the bells and whistles... for free.

[CommunityMember]

Now that hell froze over, maybe they should think about alternatives to automatically committing a timestamp update everyday.

There has been random talk for some time about how the future processes should work (including removing the daily snapshots), but the first step was moving to git and to the extent possible continuing existing processes (including those snapshots), until there is a consensus of how to move forward (since there are people who have their own external processes depending on the existing gcc processes). Once things settle, it will be time for the next discussion. Hopefully the next discussion will not take multiple years to come to a conclusion as to the path forward.

[CommunityMember]

With Stallman out of the way ....

As far as I know RMS never objected to git as a VCS (I suspect he would have preferred it be GPLv3 rather than GPLv2, but that is a nit).

[CommunityMember]

Doesn't a extremely important project such as GCC have a proper server?

They have always been about running their own infrastructure rather than using someone else's. Realistically by using svn that admittedly limited their available options, even if they were willing to consider using some other infrastructure. Now that they are using git they have a few choices if they choose to look outside themselves (github or gitlab are obvious choices, but there are others) as the server/instance just does not matter as much (everyone has a complete history cryptographically verifiable).

What about their own GitLab instance too?

They could do that, but since running infrastructure is not their goal or mission (although it may be an interest for some of the individuals), letting someone else run it has great advantages. Odds are they will continue to run their own just because they can.

[AdamOne]

With Stallman out of the way, I guess they can actually think about *convenience* for once.
Weird-ass chomo.

[timofonic]

http://gcc.gnu.org/git/?p=gcc.git;a=summary

The GitHub mirror has successfully updated: https://github.com/gcc-mirror/gcc (that has everything, including the not-fetched by default refs, plus GitHub-specific refs/pull/* refs for pull requests people created on GitHub in the past). The GitHub mirror should keep automatically updating in future.

Please use it instead of the primary mirror for cloning since gcc.gnu.org can barely handle many people doing an initial checkout.

Doesn't a extremely important project such as GCC have a proper server?

What about their own GitLab instance too?