X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=dgit.git;a=blobdiff_plain;f=dgit-maint-debrebase.7.pod;h=0f2f89288892722374aa392cab508300ae0e5824;hp=f5023e849ba5c6c271069e713a929bdf9c3974a4;hb=6afbad208f42c5f4b2f232988a7ed775b638ec7b;hpb=7ba46bbf3be9e1bb01cfa50e46ec2f67a46b6eda diff --git a/dgit-maint-debrebase.7.pod b/dgit-maint-debrebase.7.pod index f5023e84..0f2f8928 100644 --- a/dgit-maint-debrebase.7.pod +++ b/dgit-maint-debrebase.7.pod @@ -262,7 +262,7 @@ The first dgit push will require I<--overwrite>. =head1 GIT CONFIGURATION -git-debrebase does not yet support using B to merge +git-debrebase(1) does not yet support using B to merge divergent branches of development (see "OTHER MERGES" in git-debrebase(5)). You should configure git such that B does not try to merge: @@ -323,7 +323,7 @@ or if you have a working watch file =over 4 - % git debrebase new-upstream-v0 1.2.3 + % git debrebase new-upstream 1.2.3 =back @@ -400,13 +400,14 @@ git remote such as B, Note that each time you conclude a debrebase you introduce a pseudomerge into your git history, which may make it harder to read. Try to do all of the editing of the delta queue that you think will be -needed for this upload in a single debrebase, so that there is a -single debrebase stitch. +needed for this editing session in a single debrebase, so that there +is a single debrebase stitch. =head1 BUILDING AND UPLOADING You can use dpkg-buildpackage(1) for test builds. When you are ready -to build for an upload, use B. +to build for an upload, use B, B or B. Upload with B or B. Remember to pass I<--new> if the package is new in the target suite. @@ -429,9 +430,11 @@ you're using that. =head1 HANDLING DFSG-NON-FREE MATERIAL -This covers only DFSG-non-free material. Material which is legally -dangerous (for example, files which are actually illegal) cannot be -handled this way. +=head2 Illegal material + +Here we explain how to handle material that is merely DFSG-non-free. +Material which is legally dangerous (for example, files which are +actually illegal) cannot be handled this way. If you encounter possibly-legally-dangerous material in the upstream source code you should seek advice. It is often best not to make a @@ -439,22 +442,51 @@ fuss on a public mailing list (at least, not at first). Instead, email your archive administrators. For Debian that is To: dgit-owner@debian.org, ftpmaster@ftp-master.debian.org -=head2 When upstream tags releases in git +=head2 DFSG-non-free: When upstream tags releases in git + +Our approach is to maintain a DFSG-clean upstream branch, and create +tags on this branch for each release that we want to import. We then +import those tags per "Importing the release", above. -We create a DFSG-clean tag to import to master: +For the first upstream release that requires DFSG filtering: =over 4 - % git checkout -b pre-dfsg 1.2.3 + % git checkout -b upstream-dfsg 1.2.3 % git rm evil.bin % git commit -m "upstream version 1.2.3 DFSG-cleaned" % git tag -s 1.2.3+dfsg % git checkout master - % git branch -D pre-dfsg + % # proceed with "Importing the release" on 1.2.3+dfsg tag =back -=head2 When upstream releases only tarballs +And for subsequent releases (whether or not they require filtering): + +=over 4 + + % git checkout upstream-dfsg + % git merge 1.2.4 + % git rm further-evil.bin # if needed + % git commit -m "upstream version 1.2.4 DFSG-cleaned" # if needed + % git tag -s 1.2.4+dfsg + % git checkout master + % # proceed with "Importing the release" on 1.2.4+dfsg tag + +=back + +Our upstream-dfsg branch cannot be pushed to B, but since +we will need it whenever we import a new upstream version, we must +push it somewhere. Assuming that you have already set up an origin +remote per the above, + +=over 4 + + % git push --follow-tags -u origin master upstream-dfsg + +=back + +=head2 DFSG-non-free: When upstream releases only tarballs The easiest way to handle this is to add a B field to I, and a B setting in @@ -476,6 +508,10 @@ If that fails, because your branch and the NMUers work represent divergent branches of development, you have a number of options. Here we describe the two simplest. +Note that you should not try to resolve the divergent branches by +editing files in I. Changes there would either cause +trouble, or be overwritten by git-debrebase(1). + =head2 Rebasing your work onto the NMU =over 4 @@ -505,25 +541,49 @@ I<--overwrite>. Above we noted that each time you conclude a debrebase, you introduce a pseudomerge into your git history, which may make it harder to read. -A convention you can use to minimise the number of pseudomerges is to -debrebase only right before you upload. +A simple convention you can use to minimise the number of pseudomerges +is to B only right before you upload or push +to B. -Before that point, instead of editing the existing delta queue, you -append fixup commits (and reversions of commits) that alter the -upstream source to the required state. You can freely push and pull -from B during this. Just before uploading, you -debrebase, once, to tidy everything up. +It is possible, though much less convenient, to reduce the number of +pseudomerges yet further. We debrebase only (i) when importing a new +release, and (ii) right before uploading. Instead of editing the +existing delta queue, you append fixup commits (and reversions of +commits) that alter the upstream source to the required state. You +can push and pull to and from B during this. Just +before uploading, you debrebase, once, to tidy everything up. + +=head2 The debian/patches directory + +In this workflow, I is purely an output of +git-debrebase(1). You should not make changes there. They will +either cause trouble, or be ignored and overwritten by +git-debrebase(1). + +I will often be out-of-date because git-debrebase(1) +will only regenerate it when it needs to. So you should not rely on +the information in that directory. When preparing patches to forward +upstream, you should use git-format-patch(1) on git commits, rather +than sending files from I. =head2 Upstream branches -Except in the case where upstream releases only tarballs, we do not -maintain a separate 'upstream' branch (unless you also happen to be -involved in upstream development). We work with upstream tags rather -than any branches, except temporary branches used to prepare patches -for forwarding upstream, for example. +In this workflow, we specify upstream tags rather than any branches. -The thought behind this is that from Debian's point of view, upstream -releases are immutable points in history, better represented by tags. +Except when (i) upstream releases only tarballs, (ii) we require DFSG +filtering, or (iii) you also happen to be involved in upstream +development, we do not maintain any local branch corresponding to +upstream, except temporary branches used to prepare patches for +forwarding, and the like. + +The idea here is that from Debian's point of view, upstream releases +are immutable points in history. We want to make sure that we are +basing our Debian package on a properly identified upstream version, +rather than some arbitrary commit on some branch. Tags are more +useful for this. + +Upstream's branches remain available as the git remote tracking +branches for your upstream remote, e.g. I. =head2 The first ever dgit push @@ -559,17 +619,22 @@ using git-rebase(1) directly. For example, =over 4 - % git debrebase launder + % git debrebase % git rebase -i HEAD~5 # there are 4 Debian patches =back If you take this approach, you should be very careful not to start the -rebase too early. +rebase too early, +including before the most recent pseudomerge. +git-rebase without a base argument will often +start the rebase too early, +and should be avoided. +Run git-debrebase instead. =head1 SEE ALSO -dgit(1), dgit(7) +dgit(1), dgit(7), git-debrebase(1), git-debrebase(5) =head1 AUTHOR