|Version 5 (modified by simonmar, 9 years ago) (diff)|
Guidelines for using darcs with GHC
GHC uses darcs for revision control. This page describes various GHC-specific conventions for using darcs, together with some suggestions and tips for using darcs effectively.
- Try to make small patches (i.e. work in consistent increments).
- Separate changes that affect functionality from those that just affect code layout, indendation, whitespace, filenames etc. This means that when looking at patches later, we don't have to wade through loads of non-functional changes to get to the important parts of the patch.
- If possible, record often. This helps to avoid conflicts.
- Do not push conflicts (see WorkingConventions/Darcs).
- Every patch must pass at least minimal validation: see TestingPatches.
- Discuss anything you think might be controversial before pushing it.
We have a simple naming convention for certain kinds of patches:
- If your patch fixes breakage in the build, then begin the patch name with "FIX BUILD". e.g.
FIX BUILD Use the right find on Windows systems; fixes bindist creationThe point here is that if someone pulls GHC from darcs and experiences a build failure, they can try darcs pull -a -p "FIX BUILD" in order to grab patches that fix it, without grabbing anything else that might introduce further breakage.
- If your patch fixes a bug, then begin the patch name with "FIX #NNNN", where NNNN is the ticket
FIX #767 (withMVar family have a bug)
The stable branch
The preferred way to make a change to the stable branch is to first record it on the HEAD, with a note in the commit message that it should be merged. You can then also push it to the stable branch if you wish, or you can leave it for the stable branch maintainer to do.
Sometimes it is not possible to push the HEAD patch to the stable branch, as it may depend on other patches that are meant only for the HEAD. In this case the changes will need to be remade in the stable branch. The commit message for this new change should include the entire commit message of the original change, prefaced with "MERGED: ". The same patch should be pushed to both branches whenever possible, though, as it makes it easier to compare the two branches.
If you first fix a problem in the stable branch then please do apply it only to that branch rather than leaving it lying around for when you have time to make the HEAD patch.
Right now, we avoid having any conflicts in the GHC HEAD repository, because darcs currently doesn't handle conflicts well. Don't push and conflicts, even resolved ones, to the GHC HEAD. Instead amend-record or re-record your patches to fix the conflicts before pushing.
Conflicts on branches are less of a problem, because branches usually have a limited life-span. As long as the branch is not intended to be merged with the HEAD in the future (e.g. the stable branches), then small conflicts are ok.
We're aware that this policy creates problems for development branches of GHC, and this is truly unfortunate. We're hopeful that darcs' conflict handling will improve in the future and we can get the full power of darcs for separate development.
If you have permission to push patches directly to the repository (pretty easy to get, just demonstrate your competence by sending us a patch or two first), then you can use darcs push:
$ darcs push <account>@darcs.haskell.org:/home/darcs/ghc
(change ghc to the name of the repository if you're pushing changes from one of the sub-repositories, like testsuite, or a package such as base. Note: darcs push requires that SSH is working and can log in to your account on darcs.haskell.org.
Do not forget to darcs record your changes first!
Pushing a whole tree
It's good practice to have a completely clean set of repositories locally, i.e. a locally cached copy of the main repositories on darcs.haskell.org. This is useful for creating new trees quickly, or comparing your trees to the HEAD. To see which patches you have in a GHC tree relative to a clean repository, you can use the push-all script in the root of the GHC repository:
$ ./push-all --checked-out ~/ghc-HEAD --dry-run
where ~/ghc-HEAD is my vanilla HEAD, with all the sub-repositories checked out using darcs-all. This command tells me all the patches in the local repository tree relative to ~/ghc-HEAD.
To actually push to the HEAD, you can do this:
$ ./push-all email@example.com:/home/darcs
it'll use SSH for the push, but continue to use HTTP for pulling, which is what you want (HTTP is much faster than SSH for darcs operations, but for pushing we can only use SSH).