Eric S Raymond Believes Reposurgeon Is Finally Ready For Full & Correct GCC Conversion

[Delgarde]

FWIW, I don't think the end goal for reposurgeon is to improve git-svn; rather, I think the goal is to improve reposurgeon itself which has a somewhat wider scope than git-svn does:

Yeah, though not sure how useful that is... pretty much all other projects of any note have long since migrated already, using the likes of git-svn. Because by all accounts, git-svn works pretty well if you're not trying to migrate 30 years of history that's already been through several migrations before reaching the svn repo they're now trying to migrate it out of. GCC is kind of a niche case.

[timofonic]

When It's Done (TM).

Will it happen before LLVM conquers the world? Seriously, GCC needs extremely a lot more than just converting their repository to Git. It's very sad and a real shame, there's a very urgent need of competition in the world of compilers.

Some stuff I want to see:

- More reusable pieces as in LLVM.
- 900% faster programming language, new standards and platform support: Where's Rust? What about stuff such as Kotlin? What about targeting to bytecode an Emscripten-like too? What about WebAssembly?
- A lot better optimizations.
- Dramatically improving compilation speed.
- A lot better ways to check for code quality.

[coder]

I just want to know the final git repo size.

[Vistaus]

Yeah, though not sure how useful that is... pretty much all other projects of any note have long since migrated already, using the likes of git-svn. Because by all accounts, git-svn works pretty well if you're not trying to migrate 30 years of history that's already been through several migrations before reaching the svn repo they're now trying to migrate it out of. GCC is kind of a niche case.

By the time they're finally done migrating to Git, they'll start preparing to migrate to Git's successor. And so their migration circle continues...

[paulpach]

Finally Ready For Full & Correct GCC Conversion

...

It's also not quite 100% buttoned up but is said to be in the days ahead

those 2 statements contradict each other.

[CommunityMember]

When It's Done (TM).

Will it happen before LLVM conquers the world? Seriously, GCC needs extremely a lot more than just converting their repository to Git.

The conversion to git will not suddenly improve GCC development, but it will address some of the pain points that makes development more difficult than it needs to be. A number of the GCC developers reportedly already use git-svn for their day-to-day work, but as everyone who has used that method knows, that imposes some hoop jumping and limitations at times. Any change that can make GCC developers more effective is nothing but a good thing.

It's very sad and a real shame, there's a very urgent need of competition in the world of compilers.

There has almost always been competition, but the competition tends to be of the paid for (or variously licensed) variety. For corporations that is often a minor cost to the entire project, but individuals may have fewer resources when the results are not going to be producing revenue.

In many ways the historical GCC development team invited llvm to become a viable (in some ways better) alternative with their insistence of their way or the highway. So thank GCC intransigence for the llvm competition

[woife]

I still hope the other guy gets the go ahead. ESR deserves getting blown off at this point.

Why?

Because you care about what is better for gcc and its community in the long term, based on your detailed understanding of svn, git, gcc and what esr contributes?

[bearoso]

Why?

Because you care about what is better for gcc and its community in the long term, based on your detailed understanding of svn, git, gcc and what esr contributes?

I think many people are disillusioned because of ESR’s seemingly transparent pleas for hardware donations. His motives were quite suspicious and the hardware requests didn’t jive with the work needing to be done. I don’t know what the truth is, but you can’t deny that it looks sketchy.

Regardless, if the tool is complete and produces better output, then there’s no reason not to use it. Vindictiveness shouldn’t get in the way of a greater good.

[woife]

I think many people are disillusioned because of ESR’s seemingly transparent pleas for hardware donations. His motives were quite suspicious and the hardware requests didn’t jive with the work needing to be done.

I'm not sure. I have read parts of the gcc mailing list threads, and I haven't found a mail where any of his claims were shown to be untrue.
I don't know any of the internals of gcc, svn, git or any of the other involved tools (git-svn, reposurgeon, ...), but I find it easy to believe that it is indeed a complex problem domain.

I only see people that argue how "he should get the job done" here in the forum. On the actual gcc mailing list the discussion is more about whether a 100% correct transformation is worth waiting for instead of going with git-svn now. But I could not find a mailing list post where anyone would argue that reposurgeon is implementing a wrong approach.

[ssokolow]

I'm not sure. I have read parts of the gcc mailing list threads, and I haven't found a mail where any of his claims were shown to be untrue.
I don't know any of the internals of gcc, svn, git or any of the other involved tools (git-svn, reposurgeon, ...), but I find it easy to believe that it is indeed a complex problem domain.

I only see people that argue how "he should get the job done" here in the forum. On the actual gcc mailing list the discussion is more about whether a 100% correct transformation is worth waiting for instead of going with git-svn now. But I could not find a mailing list post where anyone would argue that reposurgeon is implementing a wrong approach.

I think the point is that it's fishy that any tool that's processing something like a code repository should need that much hardware when it should be perfectly possible to restructure the work into two or more streaming passes with minimal RAM requirements.