The Debate Over GCC's SVN-to-Git Conversion Approach Won't Be Settled This Year

[bug77]

If the blocker here is a few edge cases, why not convert now and keep the old repo as an upstream? After you fix the edge cases you can pull the changes and rebase (forcefully, if needed) on that.

And I'm pretty sure despite this "history is important" discussion, we're all aware going forward, nobody will ever care about some weird merge done on some GCC 4 branch years ago. On top of that, SVN allows commits in tags (because it doesn't actually have tags), would you want to pull those in as well? Because, you know, otherwise you're not preserving history.

[torsionbar28]

But that's not an even remotely accurate characterization of what either conversion is doing.

Huh? This is what YOU were suggesting as an alternative to the conversion methods.

Bullshit. Just moving to a git workflow would be a massive boon to development.

Spoken like someone who has never actually managed a project on a fractured repository.

[crystall]

You clearly haven't bothered to do any reading about this. Preserving history is obviously important.

Then why did you claim that the reason I stated for preserving it are "marketing"?

The problem is, how much effort is it worth to preserve a few, minor conversion artefacts that were broken even in the original repo?

I never commented on that. I responded to a post that claimed that preserving version history was only for "archeologists" by stating why it was important.

[torsionbar28]

The conversion is proving to be hard because the original itself is full of botched history, thanks to the sloppiness of SVN/CVS. In the (very rare) event they need the exact, original history, they can just go back and work on the SVN repo directly.

The entire purpose of this effort is to move away from SVN. Splitting the version history across two incompatible systems is worse than doing nothing at all. ESR is right - as is typically the case, after the dust settles.

[crystall]

Please spare us the ESR-style self-marketing fluff...

There is no marketing in my claims, just real world experience. I've been a Firefox developer for 8 years and we worked hard to preserve history through the various conversions and make it available through external tools because it is important. If you'll ever work on a large codebase dating back decades - and GCC is one of the oldest out there - you'll appreciate it.

The conversion is proving to be hard because the original itself is full of botched history, thanks to the sloppiness of SVN/CVS. In the (very rare) event they need the exact, original history, they can just go back and work on the SVN repo directly.

That's neither simple nor convenient and it just doesn't work if you're using automated tools.

Maxim's conversion has been ready since summer and is basically as good as it needs to be for day-to-day work. At this point, for ESR, it's just about pumping up his ego and justifying his future self-aggrandizement as a "world-renowned VCS conversion expert".

I do not know this person but I don't see why an ad-hominem attack on him would convince me that preserving history when switching VCS isn't important.

[Terrablit]

[I]

Since when did this become a competition? Just pick one, and fix any issues afterwards.

Well, it was a vaporware boast for 2 years, then Maxim came along and wrote his scripts using like 1/8 of the time, then ESR panicked and started trying to actually deliver something instead of pulling the perpetual bait and switch. They're at similar levels of completion, but Maxim's stuff has been sitting around waiting for ESR to finish being "almost done" for like half a year. And then ESR said it was totally ready to grab the spotlight again butit wasn't, so he's boasting and patronizing on the ML while the Rube Goldberg machine of indentured elves that make up the reposurgeon team is scurrying around to fix all the problems popping up. ESR's like some type of Dystopian Feudal Lord Santa that's here to deliver broken toys on others' backs. Maxim's had to apply some fixes, too, but they're quick turnarounds, and some were config changes or automated build issues.

I scanned the ML posts for my daily dose of indignation. It seems like Joseph wants to prioritize the reposurgeon thing if it can work, followed by Maxim's, followed by a third option. But that might be based on the deliverable date. I think he just wants to believe that the time wasn't wasted, and not discourage all the people ESR snagged to do his work. I think very little of ESR and his ALM / competency / work ethic (as well as his words, character and general attitude), so I don't really want to see him land this at this point because he'll boast about it again and use it to insert himself elsewhere and waste more time. But I'm not the one picking. Honestly, just having him out of my news feed would be great.

Regardless of whether Maxim's solution is picked in the end, I still think the only reason the reposurgeon solution ever picked up speed is because he threatened it with a few weeks of work. He's like some kind of ESR Jesus - Maxim POCed to save us from [ESR's] sins.

[bug77]

Shall we start placing bets on whether this happens before or after Star Citizen launches?

[stormcrow]

So ESR can brag about his "achievement" and use it to justify inserting himself into other conversion efforts in the future (perhaps with more free hardware or consulting fees).

ESR is going to do that anyway. His ego won't let him do anything less. If he gets rebuffed he'll whine about it on his blogs how he was wronged when he tried to help.

[crystall]

Why is it even so important to convert the whole history in its entirety?
Version control is primarily a tool for efficient development going forward. Archaeologists can use the old repository if desired.

Because you need to be able to search history to track down the origin of a bug, the reason why a change was introduced, to perform a bisection while looking for a regression, to be able to backport patches to older branches... I could go on for a while. For large projects having accurate and functioning history in your VCS is extremely important.

[stormcrow]

This is ridiculous at this point. Just do it already and present it as a fait accompli.