Re: HEADS-UP: new server-side validation git hook for submodule updates & call-for-help

On 18/03/2014 18:17, Johan Tibell wrote:

Lets give some example workflows for working with submodules. Here's what I think a raw (i.e. no sync-all) update to base will look like. Please correct me if I'm wrong.

# Step 1: cd ~/src/ghc/libraries/base # edit some_file git add some_file git commit -m "Commit to base repo" git push # push update to base to git.haskell.org http://git.haskell.org

I believe this doesn't work, because the normal state for a submodule is "detached HEAD", so you can't commit to it because it isn't on a branch. You have to first "get checkout master", or "git checkout -b mybranch master".

# Step 2 cd ~/src/ghc git add libraries/base git commit -m "Have GHC use the new base version" git push # push update to ghc to git.haskell.org http://git.haskell.org

Failure modes include:

* Forgetting step 2: the ghc repo will point to a slightly older base next time someone checks it out. Fixing things when in this state: just perform step 2. * Forgetting `git push` in step 1. the ghc repo will point to a base commit that doesn't exist (except on some developers machine). Fixing things when in this state: the developer who forgot to `git push` in step 1 needs to do that.

How could sync-all help us:

* sync-all push could push all repos, preventing failure case 2 above.

The second interesting workflow involving pulling new changes. This is what the raw (i.e. no sync-all) workflow will look like:

cd ~/src/ghc git pull git submodule update

Failure modes include:

* Forgetting the `submodule update` and then doing e.g. `git commit -am "some compile commit"`, reverting the pointer to e.g. base to whatever older version the developer was using. No commits are lost (nothing changes in the base repo), but the ghc repo will point to an older commit.

The other failure mode is that the submodule contains local changes, that just got overwritten by the "git submodule update". Perhaps git is better about telling you when this is about to happen and/or failing in submodule update now? What about when the submodule is on a branch? Cheers, Simon

How could sync-all help us:

* sync-all pull could always run `submodule update`.

The server-side check that Herbert added will make sure that the failure mode cannot happen, as you explicitly have to say in the commit message that you updated a submodule.

I think if base was folded into ghc.git very few people would have to deal with submodules.

On Tue, Mar 18, 2014 at 11:58 AM, Herbert Valerio Riedel mailto:hvr@gnu.org> wrote:

Hello *,

I've put in place a new server-side validation hook a few days ago, and since nobody seemed to have complained yet, I assume it didn't have any adverse effects so far :-)

It will only be triggered when Git submodule references are touched by a commit; you can find some preliminary (but incomplete) documentation and a sample session triggering validation-failure on purpose at

https://ghc.haskell.org/trac/ghc/ticket/8251#comment:4

(this will be turned into a proper wiki-page once #8251 is completed; there's some minor details wrt some corner cases that still need to be looked at)

So, this mostly addresses the server-side requirements for migrating to a proper git-submodule set-up for ghc.git;

The next steps, however, include taking care of the client-side work-flow for working with a fully "submoduled" ghc.git setup. Personally, I'm quite comfortable using direct git commands to manage such a construct, but I'm well aware not everyone is (as previous discussions here have shown). Also, as my time is rather limited, I'd like to ask interested parties to join in and help formulate the future client-side work-flow[1] and/or update (or rewrite) the 'sync-all' to provide a seamless or at least smooth transition for those GHC devs who want to keep using "sync-all" instead of using direct Git commands.

[1]: There's some difference in how tracked upstream packages and GHC-HQ owned sub-repos are to be handled workflow-wise, to avoid ending up with a noisy ghc.git history.

For instance, having ghc.git with submodules is not the same as having a huge monolithic ghc.git repository with all subrepos embedded. specifically, it might not be sensible to propagate *every* single subrepo-commit as a separate ghc.git submod-ref update, but rather in logical batches (N.B.: using submodules gives the additional ability to git bisect within subrepos instead of having to bisect always only at top-level). This is one example of things to discuss/consider when designing the new work-flow.

Cheers, hvr _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org mailto:ghc-devs@haskell.org http://www.haskell.org/mailman/listinfo/ghc-devs

_______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://www.haskell.org/mailman/listinfo/ghc-devs