Mercurial 설치와 설정
Installing Mercurial 설치를 참고하세요. Mercurial이 이미 설치되어 있어도 다음 사항을 확인하세요.
- 최신 버전을 사용하고 있는지 확인하세요
- 올바르게 설정되었는지 확인(설치페이지의 설명을 참고하세요)하세요.
어떻게...
Mozilla 코드를 어떻게 받을 수 있나요?
Gecko와 Firefox의 기본 개발 저장소를 클론받기 위해서는 아래와 같이 하세요:
hg clone http://hg.mozilla.org/mozilla-central/ src cd src
다른 저장소는 Mercurial을 이용해서 Mozilla의 소스 코드 받기를 참고하세요.
어떻게 파일을 수정하고 업데이트 하나요?
CVS처럼 작업 디렉토리에서 아무 파일이나 수정할 수 있습니다.
트리를 업데이트하는 일반적인 명령어는 다음과 같습니다:
hg pull -u
하나의 디렉토리만 업데이트 하는 일은 할 수 없습니다. 전체 트리를 업데이트 해야만 합니다.
How can I diff and patch files?
hg diff
diffs the entire tree by default. Usehg diff <dir>
to diff only the given directory. Don't forget tohg add
any new files you are adding before generating the patch.- If you are changing binary files or renaming files you may want to use git style patches with
hg diff -g
to retain that sort of information in the patch. - When applying a patch generated by Mercurial, use
patch -p1
, notpatch -p0
. (This is due to a quirk of Mercurial diff output—it refers to fictional "a" and "b" directories. Don't ask.). - If you have a git style patch with renames and/or binary changes you can use
hg import --no-commit
to apply the patch to your tree or usehg qimport
to import the patch to mq. - There's hg qimportbz extension (blog post) that lets you import patches directly from bugzilla to mq. (Since that blog post was written, the command syntax has changed. It is now:
hg qimport bz://1234567
) - If you have it,
git-apply
can also be used to apply patches. It can handle most forms of Mercurial diff output, but it won't automatically tell Mercurial about added, copied, removed or renamed files, and it cannot reverse binary changes (except to remove added binary files).
Preferred diff options are hg diff -p -U 8
which includes 8 lines of context and shows the relevant function for the block. You can make these options default in the Mercurial configuration file.
How can I generate a patch for somebody else to check-in for me?
If you don't have commit access yourself, you need to attach your patches to a bug for somebody to check in. In order to do that, you should make sure that your patch has the following conditions:
- It has a correctly formatted author name.
- It has a correctly formatted commit message.
- It is generated in git style.
Here is a very easy way to get this setup working using mq. First, you need to edit your ~/.hgrc
file. You only need to do this once. Your hgrc file should have the following contents at least:
[ui] username = John Smith <john@smith.com> [defaults] qnew = -Ue [extensions] mq = [diff] git = 1 showfunc = 1 unified = 8
Now, in order to create a patch, you should enter:
hg qnew name.patch
which opens up your favorite editor so that you can enter a good commit message. A good sample commit message looks like: "Bug 123456 - Change this thing to work better by doing something; r=reviewers". You don't need to enter the final commit message (for example, if you haven't received reviews yet). You can edit the commit message of the current mq
patch at any time using hg qref -e
.
The ~/.hgrc
default -U
argument for the qnew command add the author information to the patch (using a From:
header), and the default -e
argument opens up the editor for you to type in a commit message.
Now, after editing some source code, you can use hg qref
to update your patch file.
To attach the patch to the bug, you can use the file named name.patch
located in your <repository>/.hg/patches
directory directly.
어떻게 변경 내용을 적용(check in)하나요?
필수 설정
변경 내용을 커밋하기 전에 이 내용을 ~/.hgrc
파일에 추가해야 합니다:
[ui] username = Your Name <your.email@example.com>
mozilla-central이나 다른 mozilla-hosted 저장소에 변경내용을 올리기 위해서는 커밋 권한을 가지고 있어야 하고 (your-local-hg-root)/.hg/hgrc
(~/.hgrc
파일이 아닙니다) 파일에 다음의 내용을 넣어야 합니다:
[paths] default = http://hg.mozilla.org/mozilla-central/ default-push = ssh://hg.mozilla.org/mozilla-central/
또한 ssh에 hg.mozilla.org에 접속할 사용자 이름이 무엇인지 알려주어야 합니다. 이 이름은 Mozilla LDAP 계정과 연결된 사용자 이름이어야 합니다. 이를 위해서 위에 있는 default-push 경로는 좀 더 복잡하게 만들거나 (user@host.name@hg.mozilla.org
) ~/.ssh/config에 다음 내용을 넣어서 해결할 수 있습니다:
Host hg.mozilla.org User user@host.domain
커밋하려고 하는 내용 확인
다음으로 정말로 커밋하고자 하는 내용을 커밋하게 될 것인지 확인합니다(특히나 다른 까다로운 커밋을 머지할 때에 중요합니다):
hg status hg diff
status
명령어는 작업 디렉토리의 변경된 내용이 있는 파일과 저장소(부모 리비전으로서 hg parents
를 이용해서 볼 수 있습니다)에 있는 내용을 비교해서 보여줍니다. hg diff
는 이 파일들의 실제 변경 내역을 하나의 diff로 보여줍니다. 더 자세한 내용을 보기 위해서 -U8 옵션을 설정할 수 있습니다.
로컬 저장소에 커밋
다음 단계로 변경내용을 로컬 저장소에 커밋합니다:
hg commit
이렇게 하면 hg status
로 보고된 변경내용을 매번 커밋합니다. hg commit filenames
을 이용해서 특정 파일이나 디렉토리만 커밋 할 수 있습니다. hg commit -u "Someone Else <someone@example.com>"를 사용해서 다른사람을 대신해서 커밋할 수도 있습니다
. 더 자세한 내용은 hg help commit
을 살펴보세요.
새로운 파일을 저장소에 추가된고 필요없는 파일을 지우려면 다음 명령어를 사용하세요.
hg addremove
푸시하려는 내용 확인
푸시를 하기 전에 어떠한 변경내용(changeset)이 푸시될지 확인할 수 있습니다. 이를 위해서 outgoing
명령어를 사용합니다:
hg outgoing
이렇게 하면 기본 원격 저장소에 푸시할 변경내용의 목록을 보여줍니다. 다른 저장소에 푸시될 변경내용을 확인하려면 저장소 아규먼트를 추가합니다.
중앙 저장소에 푸시
중앙 저장소에 이 변경사항을들 푸시해서 올립니다:
hg push
실수로 푸시하는 일 방지
qpush
명령어가 쉽게 push
로 잘못 입력될 수 있기 때문에, MQ 패치를 적용할 때 실수로 푸시를 하게될 수 있습니다. 실수로 푸시하는 일을 방지하기 위해서는 아래의 내용을 ~/.hgrc
파일에 기입합니다:
[hooks] pre-push = read -p 'Are you sure you want to push to remote? (y/n): '; if test "$REPLY" != "y"; then echo 'Push cancelled'; exit 1; fi
이렇게 하면 push
명령어를 입력할 때 마다 사용자가 확인을 해야 하고 y 이외의 다른 내용을 입력하면 푸시가 취소됩니다.
How do I deal with "abort: push creates new remote heads!"?
Whatever you do, don't do 'push -f' like the message suggests. (It'll probably fail anyway, but don't try it.)
Someone pushed new commits upstream since your last pull. You then committed your patch locally and are trying to push that commit upstream; that upstream has a different tip commit that you started from.
YOU ---> o 9123b7791b52 - Kaitlin Jones <kaitlin@example.net> - Bug 123456 - Trebled fromps (r=gavin) | TRUNK ---> | o 306726089e22 - Robert Longson <longsonr@example.com> - Bug 437448. New-style nsSVGString | | | o ba9b9a7c52a5 - Robert Longson <longsonr@example.com> - Bug 437448. New-style nsSVGString |/ o f8f4360bf155 - Robert O'Callahan <robert@example.org> - Bug 421436. Remove hack that gives
There are three things you can do. In all cases, you are strongly encouraged to take steps to back up your patch (perhaps by obtaining a diff: hg log -p -r 12345 to show the patch for rev 12345).
Using hg merge
Run hg pull
, then hg merge
. If there are any merge conflicts, hg will open a merge program to try to help you resolve them manually. (If you fail to make sense of the merge tool, that's OK. Just close it; hg will detect that the conflicts weren't all resolved and spit out some hg update -C
commands that you can use to undo and then retry the busted merge.)
Even if there were no conflicts, you have to commit the merge: hg commit -m "Merge bug 123456"
.
YOU ---> o 1fe7659c29a9 - Kaitlin Jones <kaitlin@example.net> - Merge bug 123456. |\ o | 9123b7791b52 - Kaitlin Jones <kaitlin@example.net> - Bug 123456 - Trebled fromps (r=gavin) | | TRUNK ---> | o 306726089e22 - Robert Longson <longsonr@example.com> - Bug 437448. New-style nsSVGString | | | o ba9b9a7c52a5 - Robert Longson <longsonr@example.com> - Bug 437448. New-style nsSVGString |/ o f8f4360bf155 - Robert O'Callahan <robert@example.org> - Bug 421436. Remove hack that gives
Now you can hg push
as normal.
This leaves a merge commit in the log, which some people find annoying, so it's usually better to avoid this solution.
If you decide to use this method, it is advantageous to enable the Mercurial fetch
extension (by means of a fetch =
line in the [extensions]
section of your <repository>/.hg/hgrc
or your $HOME/.hgrc
file) so you can do the pull + merge + commit
sequence in one step (assuming no merge conflicts) by using the hg fetch
command.
Using Mercurial queues
A way to solve your problem without leaving a merge commit is to import your commit into mq control, pop it off the queue, update, and then commit it again before pushing:
% hg log -l 5 415[tip] d1accb6ee840 2008-04-30 09:57 -0700 vladimir b=430873; fast path drawImage with a canvas as source 414 3a3ecbb4873e 2008-04-30 09:55 -0700 vladimir cvs sync % hg qimport -r 415 Turn the new commit you made into a MQ patch % hg qtop 415.diff % hg qpop Patch queue now empty Pop the MQ patch off the stack. At this point, the tip of your tree is also in the remote repository % hg pull -u ... various pull/update messages ... % hg qpush applying 415.diff Now at: 415.diff Fix up conficts as necessary; if you fixed up any, run hg qrefresh first % hg incoming comparing with ssh://hg.mozilla.org/mozilla-central/ searching for changes no changes found Make sure nobody pushed while you were merging. Otherwise, qpop, pull and qpush again % hg qfinish 415.diff Turn the MQ patch into a regular revision % hg log -l 5 verify that the log looks good, with your commit on top % hg push
If you already use mq to manage your patches, then just make sure you pull/update right before committing and pushing your patch. If you have problems with the above, it's ok to just do a simple merge as described previously.
In this case, if there are merge conflicts, MQ will produce .rej
files. Some people prefer this over being thrown into a merge program.
Using hg rebase
(Requires Mercurial 1.1 or higher)
This is the simplest and easiest way to deal with the problem. You can rebase your local changesets (including mq patches) on top of the tip using the rebase extension. To do this, simply enable the extension by adding this to the following section of your .hgrc
:
[extensions] hgext.rebase =
Now, type hg pull --rebase to pull and rebase your local changesets at the same time.
More details and options can be found at the Mercurial wiki.
If you have conflicting changes, you'll be thrown into your preferred merge tool.
hg pull
without --rebase
after your hg commit
, you will have to first undo that by doing hg rollback
. Only the very last action can be undone, so if you did anything else after the hg pull
, you're out of luck.hg diff > old.diff hg revert . hg pull --rebase hg push patch -p 1 -i old.diff rm old.diff
How do I see what these commands will do before I do them?
If you want to see what hg commit
will commit, run hg diff
first.
If you want to see what hg push
will push, run hg outgoing
first.
If you want to see what hg pull
will pull, run hg incoming
first.
These pairs of commands all take similar arguments, for good reason. These are a good idea to use all the time when you're learning mercurial. And even once you're an expert, always doing outgoing before push is a good idea.
How can I customize the format of the patches Mercurial creates?
Edit your ~/.hgrc
file and add some lines like these:
[defaults] diff=-U 8 -p qdiff=-U 8 #for Mercurial 1.1 use: #qdiff=-U 8 -p
[diff] git=true
The [defaults]
section affects the default output of the hg diff
and hg qdiff
commands. The options behave the same as they would on the command line. Try hg help diff
for more info.
The [diff]
section affects all patches generated by Mercurial, even hg export
and those generated for Mercurial's internal use. You need to be a lot more careful with this, but git=true
is recommended. Without it, Mercurial cannot diff binary files and does not track the execute mode bit. You may want to add showfunc=true
here to get diffs that show the function being changed in each hunk, and you may want to add unified=8
to make all diffs (including the internal ones mq uses) have 8 lines of context. Note that this may increase the risk of mq patches failing to apply!
How do I get a diff -w?
There is a known issue where hg diff -w
doesn't work.
To get around this add the following to your ~/.hgrc
file, editing existing sections where applicable:
[extensions] hgext.extdiff = [extdiff] cmd.diffw = diff opts.diffw = -w -r -N -p -U 8
You can, of course, add your other favourite diff options to opts.diffw
. Once you've added this, you will now have a new hg command, hg diffw
.
hg diffw -r -2 is the equivalent of hq qdiff -w for the topmost patch in the queue.
How do I generate a bundle?
Sometimes the tree will be under sheriff control, and they will specifically ask for a bundle. If you don't have sufficient rights to push to mozilla-central, you might also generate bundles and attach them to bugs when you add checkin-needed to a bug after it has the necessary reviews and approval.
Creating a bundle is quite simple. Once you have your changes all done, commit them to your local repository. You can also commit more than one changeset. Once you have everything committed, instead of pushing you'll run hg bundle
:
hg bundle outputfile.hg
By default, hg bundle
will bundle everything that hg outgoing
would display (and what you would push with hg push
). You can limit how far forward you want to bundle as well by specifying a revision with -r
. That will take all changes up until that revision.
Backing out changes
Reverting the whole tree to a known-good revision
It's easy, like using a sledgehammer is easy. But this is usually overkill.
$ hg pull -u $ hg revert --all -r a0193d83c208 # use your known-good revision id here $ hg commit # be kind, include the revision id in your commit message $ hg push
There's a more precise alternative:
Backing out a single changeset
Suppose changeset f8f4360bf155
broke something.
$ hg pull -u $ hg backout f8f4360bf155 # use the revision id of the bad change here
This creates and commits a new changeset that reverts all the changes in that revision.
If you see this message:
the backout changeset is a new head - do not forget to merge
That means you need to merge, because your history now looks like this:
YOU ---> o 9123b7791b52 - Kaitlin Jones <kaitlin@example.net> - Backed out changeset f8f4360bf155 | TRUNK ---> | o 4e5bfb83643f - Simon Montagu <smontagu@example.org> - imported patch 435856 | | | o 6ee23de41631 - Phil Ringnalda <philringnalda@example.com> - Bug 438526 - Opening links w | | | o 22baa05d0e8a - Robert O'Callahan <robert@example.org> - Remove DOM testcase from exclusi | | | o c1aec2094f7e - Robert Longson <longsonr@example.com> - Bug 437448. New-style nsSVGString | | | o 306726089e22 - Robert Longson <longsonr@example.com> - Bug 437448. New-style nsSVGString | | | o ba9b9a7c52a5 - Robert Longson <longsonr@example.com> - Bug 437448. New-style nsSVGString |/ o f8f4360bf155 - Robert O'Callahan <robert@example.org> - Bug 421436. Remove hack that gives | ⋮ (the past)
Your backout changeset is based on an old revision. It doesn't contain more recent changes.
Handle this like any other merge. If you've never done a merge before, get help. (It could be trivial or it could be a huge headache. There's plenty about merging elsewhere in this FAQ.)
Backing out multiple changesets that are not at tip
Your push of many changesets was bad, but you didn't find out until after a lot of subsequent activity. You want to back out the bad but keep the probably-good changesets from the subsequent activity.
NB: this is hard, error-prone, and will likely b0rk your local tree if you mess up. If in doubt, ask someone else to do it for you.
What one would really like to do is
hg backout --from-rev $FIRST_BAD --to-rev $LAST_BAD
or, if local hg could access pushlog
hg backout --push $LAST_BAD
but, since hg can't do either (yet), we're stuck. What we'll do instead is travel back in hg-time to the last bad cset, revert all changes between last-bad and last-good, then merge that reversion to the current tip. Your tree might look like
MERGE_TO ---> @ c6abfc89215a - Chris Pearce <chris@pearce.org.nz> - Bug 485288 - Replace all usage of video autobuffer attribute with preload=auto. a=test-fix | o d6e8fddeb95b - Chris Pearce <chris@pearce.org.nz> - Bug 548523 - Disable test_preload_actions.html case 9 until bug 568402 is fixed. a=test-fix | o 571d2664ead0 - Chris Pearce <chris@pearce.org.nz> - Bug 548523 - Don't show throbber on video controls if we're not loading a resource. r=dolske a=blocking2.0 | o 3f7dfabab5e4 - Rich Dougherty <rich@rd.gen.nz>, Chris Pearce <chris@pearce.org.nz> - Bug 548523 - Replace HTMLMediaElement.autobuffer attribute with 'preload'. r=roc a=blocking2.0 | o d7d9cf4ab76a - Chris Pearce <chris@pearce.org.nz> - Bug 519897 - Supported indexed Ogg files. r=doublec | o 2a0e5811ece9 - Chris Pearce <chris@pearce.org.nz> - Commit merge of backout of 66dcf25705f9. a=backout |\ | o 958a30df30dd - Chris Pearce <chris@pearce.org.nz> - Backed out changeset 66dcf25705f9 | | o | 6eead86e13dd - Michael Wu <mwu@mozilla.com> - Bug 556644 - Yet another removed-files.in update, r=rs a=blocking-beta5 | | o | 69c6ce104f45 - Chris Jones <jones.chris.g@gmail.com> - Bug 588216: Avoid race between IO-thread loop->PostTask() and main-thread loop->SetNestableTasksAllowed() that led to Tasks being ignored. r=bent | | o | 73899b33ca4b - Ben Turner <bent.mozilla@gmail.com> - Bug 576716 - 'Crash [@ TypedArrayTemplate<int>::init] or [@ TypedArrayTemplate<int>::create]'. Adding a test, r=vlad a=blocker | | o | 3bd62d459019 - Mark Banner <bugzilla@standard8.plus.com> - Follow up to bug 587984 bustage fix for l10n repacks. r=Pike,ted,a=bustage-fix | | o | e1ca9091e5a6 - Benjamin Stover <webapps@stechz.com> - Bug 575731: Make test more stable in the face of various themes. r,a=sicking | | o | bb200e1f52b4 - Jonas Sicking <jonas@sicking.cc> - Bug 550959: Require version 2.5 of python. r=ted a=sicking | | o | 992491c618de - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (12/12): fix assertions in nsStyleAnimation triggered by part 3. r=dbaron a2.0=dbaron | | o | 2f078585a0f6 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (11/12): Make all assertions fatal in Declaration.h, Declaration.cpp, nsCSSDataBlock.h, nsCSSDataBlock.cpp, nsCSSValue.h, nsCSSValue.cpp, nsCSSProps.h, and nsCSSProps.cpp. r=dbaron a2.0=dbaron | | o | 5a9bd15fd7a8 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (10/12): Don't directly manipulate the contents of mTempData in the CSS parser. r=dbaron a2.0=dbaron | | o | 4bb2e0074aeb - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (9/12): Add an AddLonghandProperty method to nsCSSExpandedDataBlock. r=dbaron a2.0=dbaron | | o | 659a0864e035 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (8/12): remove the last MoveValue call from the CSS parser. r=dbaron a2.0=dbaron | | o | 980f0170d982 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (7/12): cleanup pass on css/Declaration.{h,cpp} and nsCSSDataBlock.{h,cpp}. r=dbaron a2.0=dbaron | | o | f09c1638d3c1 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (6/12): remove vestiges of nsCSSType. r=dbaron a2.0=dbaron | | o | b88472b0af90 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (5/12): eliminate ValueList as a storage type. r=dbaron a2.0=dbaron | | o | a3e21759b570 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (4/12): eliminate ValuePairList as a storage type. r=dbaron a2.0=dbaron | | o | ed89c9e297ab - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (3/12): eliminate Rect as a storage type. r=dbaron a2.0=dbaron | | o | 4fc85e572c38 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (2/12): eliminate ValuePair as a storage type. r=dbaron a2.0=dbaron | | o | 301875d4f9b6 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (1/12): Move all the CSS 'storage types' (rect, value pair, etc) to nsCSSValue.h and their code to nsCSSValue.cpp. r=dbaron a2.0=dbaron | | LAST_BAD ---> o | 90ad165ae21b - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part i: Use nsIWidget::CreateChild in nsIView::CreateWidget* (where possible). r=roc a=blocking-fennecb1 | | o | 037a5d6b376a - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part h: Add an nsIWidget::CreateChild interface to sweep (relevant to this bug) code dealing with native widgets under the widget/src/* rug. sr=roc | | o | f1af117d4598 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part g: Split nsIView::CreateWidget into CreateWidget, CreateWidgetForParent, and CreateWidgetForPopup in preparation of eliminating IIDs here. sr=roc | | o | 5bf0b315a5aa - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part f: Split out window initialization code in preparation for multiple CreateWidget* methods. r=roc | | o | 353da995af6f - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part e: Simplify the logic for creating popup widgets. r=roc | | o | 7735c00eabe9 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part d: Simplify nsView::LoadWidget and return early if it fails. r=roc | | o | 7b17bcefb174 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part c: Initialize default widget init data earlier so that it's always available. r=roc | | o | c5945b6a97ed - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part b: Remove nsIDeviceContext::SupportsNativeWidgets because it's not used meaningfully, and will be confusing in content processes. sr=roc | | o | 5452db293694 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part a: Add nsIView::Impl() and nsView::CreateWidget() to get rid of |static_cast<nsview*>(this)|. r=roc | | o | 9407827b5166 - Chris Jones <jones.chris.g@gmail.com> - Bug 582075, part 0.5: Add support for aInitData=NULL to the Windows nsWindow implementation. r=dougt | | o | 88a279b64473 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part 0: Log the repaint region bounding rect in DumpPaintEvent. r=roc | | o | cebb111fbfc4 - Robert O'Callahan <roc@ocallahan.org> - Bug 585817. Part 3: Remove nsSVGUtils::GetThebesComputationalSurface and use gfxPlatform::ScreenReferenceSurface instead. r=jwatt | | o | 7b3726c3a580 - Robert O'Callahan <roc@ocallahan.org> - Bug 585817. Part 2: Change nsIPresShell::CreateRenderingContext to GetReferenceRenderingContext, that uses the shared 1x1 surface, and use it all over the place. r=mats,sr=dbaron | | FIRST_BAD ---> o | b3e968d831ec - Robert O'Callahan <roc@ocallahan.org> - Bug 585817. Part 1: Create a single static 1x1 surface in gfxPlatform that can be used to create contexts for text measurement etc. r=vlad | | LAST_GOOD ---> o | 55ef0e0529bc - Mounir Lamouri <mounir.lamouri@gmail.com> - Bug 506554 - Implement the CSS3 pseudo-classes :required and :optional for HTML. r=sicking sr=bz a2.0=blocking | | </int></int>
We want to erase all changes between $FIRST_BAD
and $LAST_BAD
, then merge to $MERGE_TO
$ MERGE_TO="c6abfc89215a" $ LAST_BAD="90ad165ae21b" $ LAST_GOOD="55ef0e0529bc" $ hg up -r $LAST_BAD 90 files updated, 0 files merged, 4 files removed, 0 files unresolved
We just traveled back in time to the last cset we want to nuke. Now we'll revert the changes between $FIRST_BAD
and $LAST_BAD
.
$ hg revert --all --no-backup -r $LAST_GOOD reverting accessible/src/msaa/nsTextAccessibleWrap.cpp reverting content/svg/content/src/SVGMotionSMILPathUtils.h reverting content/svg/content/src/nsSVGPathElement.cpp reverting gfx/src/nsIDeviceContext.h reverting gfx/src/thebes/nsThebesDeviceContext.cpp reverting gfx/src/thebes/nsThebesDeviceContext.h reverting gfx/thebes/gfxPlatform.cpp reverting gfx/thebes/gfxPlatform.h reverting layout/base/nsCSSFrameConstructor.cpp reverting layout/base/nsDocumentViewer.cpp reverting layout/base/nsIPresShell.h reverting layout/base/nsPresShell.cpp reverting layout/build/nsLayoutStatics.cpp reverting layout/generic/nsFrameFrame.cpp reverting layout/generic/nsObjectFrame.cpp reverting layout/generic/nsSimplePageSequence.cpp reverting layout/generic/nsTextFrameThebes.cpp reverting layout/inspector/src/inFlasher.cpp reverting layout/printing/nsPrintEngine.cpp reverting layout/svg/base/src/nsSVGForeignObjectFrame.cpp reverting layout/svg/base/src/nsSVGGlyphFrame.cpp reverting layout/svg/base/src/nsSVGImageFrame.cpp reverting layout/svg/base/src/nsSVGPathGeometryFrame.cpp reverting layout/svg/base/src/nsSVGUtils.cpp reverting layout/svg/base/src/nsSVGUtils.h reverting layout/xul/base/src/nsMenuPopupFrame.cpp reverting layout/xul/base/src/nsSplitterFrame.cpp reverting layout/xul/base/src/tree/src/nsTreeBodyFrame.cpp reverting view/public/nsIView.h reverting view/src/nsView.cpp reverting view/src/nsView.h reverting widget/public/nsIWidget.h reverting widget/src/beos/nsWindow.h reverting widget/src/cocoa/nsChildView.h reverting widget/src/cocoa/nsCocoaWindow.h reverting widget/src/windows/nsWindow.cpp reverting widget/src/xpwidgets/nsBaseWidget.cpp reverting widget/src/xpwidgets/nsBaseWidget.h $ hg commit -m 'Back out bug 585817 and bug 582057' created new head
After committing the reversion, we now have a new head that doesn't include the probably-good changes after $LAST_BAD
.
@ 1767c1fb9418 - Chris Jones <jones.chris.g@gmail.com> - Back out bug 585817 and bug 582057 | | o c6abfc89215a - Chris Pearce <chris@pearce.org.nz> - Bug 485288 - Replace all usage of video autobuffer attribute with preload=auto. a=test-fix | | | o d6e8fddeb95b - Chris Pearce <chris@pearce.org.nz> - Bug 548523 - Disable test_preload_actions.html case 9 until bug 568402 is fixed. a=test-fix | | | o 571d2664ead0 - Chris Pearce <chris@pearce.org.nz> - Bug 548523 - Don't show throbber on video controls if we're not loading a resource. r=dolske a=blocking2.0 | | | o 3f7dfabab5e4 - Rich Dougherty <rich@rd.gen.nz>, Chris Pearce <chris@pearce.org.nz> - Bug 548523 - Replace HTMLMediaElement.autobuffer attribute with 'preload'. r=roc a=blocking2.0 | | | o d7d9cf4ab76a - Chris Pearce <chris@pearce.org.nz> - Bug 519897 - Supported indexed Ogg files. r=doublec | | | o 2a0e5811ece9 - Chris Pearce <chris@pearce.org.nz> - Commit merge of backout of 66dcf25705f9. a=backout | |\ | | o 958a30df30dd - Chris Pearce <chris@pearce.org.nz> - Backed out changeset 66dcf25705f9 | | | | o | 6eead86e13dd - Michael Wu <mwu@mozilla.com> - Bug 556644 - Yet another removed-files.in update, r=rs a=blocking-beta5 | | | | o | 69c6ce104f45 - Chris Jones <jones.chris.g@gmail.com> - Bug 588216: Avoid race between IO-thread loop->PostTask() and main-thread loop->SetNestableTasksAllowed() that led to Tasks being ignored. r=bent | | | | o | 73899b33ca4b - Ben Turner <bent.mozilla@gmail.com> - Bug 576716 - 'Crash [@ TypedArrayTemplate<int>::init] or [@ TypedArrayTemplate<int>::create]'. Adding a test, r=vlad a=blocker | | | | o | 3bd62d459019 - Mark Banner <bugzilla@standard8.plus.com> - Follow up to bug 587984 bustage fix for l10n repacks. r=Pike,ted,a=bustage-fix | | | | o | e1ca9091e5a6 - Benjamin Stover <webapps@stechz.com> - Bug 575731: Make test more stable in the face of various themes. r,a=sicking | | | | o | bb200e1f52b4 - Jonas Sicking <jonas@sicking.cc> - Bug 550959: Require version 2.5 of python. r=ted a=sicking | | | | o | 992491c618de - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (12/12): fix assertions in nsStyleAnimation triggered by part 3. r=dbaron a2.0=dbaron | | | | o | 2f078585a0f6 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (11/12): Make all assertions fatal in Declaration.h, Declaration.cpp, nsCSSDataBlock.h, nsCSSDataBlock.cpp, nsCSSValue.h, nsCSSValue.cpp, nsCSSProps.h, and nsCSSProps.cpp. r=dbaron a2.0=dbaron | | | | o | 5a9bd15fd7a8 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (10/12): Don't directly manipulate the contents of mTempData in the CSS parser. r=dbaron a2.0=dbaron | | | | o | 4bb2e0074aeb - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (9/12): Add an AddLonghandProperty method to nsCSSExpandedDataBlock. r=dbaron a2.0=dbaron | | | | o | 659a0864e035 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (8/12): remove the last MoveValue call from the CSS parser. r=dbaron a2.0=dbaron | | | | o | 980f0170d982 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (7/12): cleanup pass on css/Declaration.{h,cpp} and nsCSSDataBlock.{h,cpp}. r=dbaron a2.0=dbaron | | | | o | f09c1638d3c1 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (6/12): remove vestiges of nsCSSType. r=dbaron a2.0=dbaron | | | | o | b88472b0af90 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (5/12): eliminate ValueList as a storage type. r=dbaron a2.0=dbaron | | | | o | a3e21759b570 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (4/12): eliminate ValuePairList as a storage type. r=dbaron a2.0=dbaron | | | | o | ed89c9e297ab - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (3/12): eliminate Rect as a storage type. r=dbaron a2.0=dbaron | | | | o | 4fc85e572c38 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (2/12): eliminate ValuePair as a storage type. r=dbaron a2.0=dbaron | | | | o | 301875d4f9b6 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (1/12): Move all the CSS 'storage types' (rect, value pair, etc) to nsCSSValue.h and their code to nsCSSValue.cpp. r=dbaron a2.0=dbaron |/ / o | 90ad165ae21b - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part i: Use nsIWidget::CreateChild in nsIView::CreateWidget* (where possible). r=roc a=blocking-fennecb1 | | o | 037a5d6b376a - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part h: Add an nsIWidget::CreateChild interface to sweep (relevant to this bug) code dealing with native widgets under the widget/src/* rug. sr=roc | | o | f1af117d4598 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part g: Split nsIView::CreateWidget into CreateWidget, CreateWidgetForParent, and CreateWidgetForPopup in preparation of eliminating IIDs here. sr=roc | | o | 5bf0b315a5aa - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part f: Split out window initialization code in preparation for multiple CreateWidget* methods. r=roc | | o | 353da995af6f - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part e: Simplify the logic for creating popup widgets. r=roc | | o | 7735c00eabe9 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part d: Simplify nsView::LoadWidget and return early if it fails. r=roc | | o | 7b17bcefb174 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part c: Initialize default widget init data earlier so that it's always available. r=roc | | o | c5945b6a97ed - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part b: Remove nsIDeviceContext::SupportsNativeWidgets because it's not used meaningfully, and will be confusing in content processes. sr=roc | | o | 5452db293694 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part a: Add nsIView::Impl() and nsView::CreateWidget() to get rid of |static_cast<nsview*>(this)|. r=roc | | o | 9407827b5166 - Chris Jones <jones.chris.g@gmail.com> - Bug 582075, part 0.5: Add support for aInitData=NULL to the Windows nsWindow implementation. r=dougt | | o | 88a279b64473 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part 0: Log the repaint region bounding rect in DumpPaintEvent. r=roc | | o | cebb111fbfc4 - Robert O'Callahan <roc@ocallahan.org> - Bug 585817. Part 3: Remove nsSVGUtils::GetThebesComputationalSurface and use gfxPlatform::ScreenReferenceSurface instead. r=jwatt | | o | 7b3726c3a580 - Robert O'Callahan <roc@ocallahan.org> - Bug 585817. Part 2: Change nsIPresShell::CreateRenderingContext to GetReferenceRenderingContext, that uses the shared 1x1 surface, and use it all over the place. r=mats,sr=dbaron | | o | b3e968d831ec - Robert O'Callahan <roc@ocallahan.org> - Bug 585817. Part 1: Create a single static 1x1 surface in gfxPlatform that can be used to create contexts for text measurement etc. r=vlad | | o | 55ef0e0529bc - Mounir Lamouri <mounir.lamouri@gmail.com> - Bug 506554 - Implement the CSS3 pseudo-classes :required and :optional for HTML. r=sicking sr=bz a2.0=blocking | | </int></int>
We want to merge with those probably-good changesets. Note that if any of those probably-good changesets touched code you backed out, you'll need to resolve merge conflicts.
$ hg merge $MERGE_TO 92 files updated, 0 files merged, 2 files removed, 0 files unresolved (branch merge, don't forget to commit) $ hg commit -m 'Merge backout'
OK, good to go. Your tree should look something like
@ 8fccc434c295 - Chris Jones <jones.chris.g@gmail.com> - Merge backout |\ | o 1767c1fb9418 - Chris Jones <jones.chris.g@gmail.com> - Back out bug 585817 and bug 582057 | | o | c6abfc89215a - Chris Pearce <chris@pearce.org.nz> - Bug 485288 - Replace all usage of video autobuffer attribute with preload=auto. a=test-fix | | o | d6e8fddeb95b - Chris Pearce <chris@pearce.org.nz> - Bug 548523 - Disable test_preload_actions.html case 9 until bug 568402 is fixed. a=test-fix | | o | 571d2664ead0 - Chris Pearce <chris@pearce.org.nz> - Bug 548523 - Don't show throbber on video controls if we're not loading a resource. r=dolske a=blocking2.0 | | o | 3f7dfabab5e4 - Rich Dougherty <rich@rd.gen.nz>, Chris Pearce <chris@pearce.org.nz> - Bug 548523 - Replace HTMLMediaElement.autobuffer attribute with 'preload'. r=roc a=blocking2.0 | | o | d7d9cf4ab76a - Chris Pearce <chris@pearce.org.nz> - Bug 519897 - Supported indexed Ogg files. r=doublec | | o | 2a0e5811ece9 - Chris Pearce <chris@pearce.org.nz> - Commit merge of backout of 66dcf25705f9. a=backout |\ \ | o | 958a30df30dd - Chris Pearce <chris@pearce.org.nz> - Backed out changeset 66dcf25705f9 | | | o | | 6eead86e13dd - Michael Wu <mwu@mozilla.com> - Bug 556644 - Yet another removed-files.in update, r=rs a=blocking-beta5 | | | o | | 69c6ce104f45 - Chris Jones <jones.chris.g@gmail.com> - Bug 588216: Avoid race between IO-thread loop->PostTask() and main-thread loop->SetNestableTasksAllowed() that led to Tasks being ignored. r=bent | | | o | | 73899b33ca4b - Ben Turner <bent.mozilla@gmail.com> - Bug 576716 - 'Crash [@ TypedArrayTemplate<int>::init] or [@ TypedArrayTemplate<int>::create]'. Adding a test, r=vlad a=blocker | | | o | | 3bd62d459019 - Mark Banner <bugzilla@standard8.plus.com> - Follow up to bug 587984 bustage fix for l10n repacks. r=Pike,ted,a=bustage-fix | | | o | | e1ca9091e5a6 - Benjamin Stover <webapps@stechz.com> - Bug 575731: Make test more stable in the face of various themes. r,a=sicking | | | o | | bb200e1f52b4 - Jonas Sicking <jonas@sicking.cc> - Bug 550959: Require version 2.5 of python. r=ted a=sicking | | | o | | 992491c618de - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (12/12): fix assertions in nsStyleAnimation triggered by part 3. r=dbaron a2.0=dbaron | | | o | | 2f078585a0f6 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (11/12): Make all assertions fatal in Declaration.h, Declaration.cpp, nsCSSDataBlock.h, nsCSSDataBlock.cpp, nsCSSValue.h, nsCSSValue.cpp, nsCSSProps.h, and nsCSSProps.cpp. r=dbaron a2.0=dbaron | | | o | | 5a9bd15fd7a8 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (10/12): Don't directly manipulate the contents of mTempData in the CSS parser. r=dbaron a2.0=dbaron | | | o | | 4bb2e0074aeb - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (9/12): Add an AddLonghandProperty method to nsCSSExpandedDataBlock. r=dbaron a2.0=dbaron | | | o | | 659a0864e035 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (8/12): remove the last MoveValue call from the CSS parser. r=dbaron a2.0=dbaron | | | o | | 980f0170d982 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (7/12): cleanup pass on css/Declaration.{h,cpp} and nsCSSDataBlock.{h,cpp}. r=dbaron a2.0=dbaron | | | o | | f09c1638d3c1 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (6/12): remove vestiges of nsCSSType. r=dbaron a2.0=dbaron | | | o | | b88472b0af90 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (5/12): eliminate ValueList as a storage type. r=dbaron a2.0=dbaron | | | o | | a3e21759b570 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (4/12): eliminate ValuePairList as a storage type. r=dbaron a2.0=dbaron | | | o | | ed89c9e297ab - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (3/12): eliminate Rect as a storage type. r=dbaron a2.0=dbaron | | | o | | 4fc85e572c38 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (2/12): eliminate ValuePair as a storage type. r=dbaron a2.0=dbaron | | | o---+ 301875d4f9b6 - Zack Weinberg <zweinberg@mozilla.com> - Bug 576044 (1/12): Move all the CSS 'storage types' (rect, value pair, etc) to nsCSSValue.h and their code to nsCSSValue.cpp. r=dbaron a2.0=dbaron / / | o 90ad165ae21b - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part i: Use nsIWidget::CreateChild in nsIView::CreateWidget* (where possible). r=roc a=blocking-fennecb1 | | | o 037a5d6b376a - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part h: Add an nsIWidget::CreateChild interface to sweep (relevant to this bug) code dealing with native widgets under the widget/src/* rug. sr=roc | | | o f1af117d4598 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part g: Split nsIView::CreateWidget into CreateWidget, CreateWidgetForParent, and CreateWidgetForPopup in preparation of eliminating IIDs here. sr=roc | | | o 5bf0b315a5aa - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part f: Split out window initialization code in preparation for multiple CreateWidget* methods. r=roc | | | o 353da995af6f - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part e: Simplify the logic for creating popup widgets. r=roc | | | o 7735c00eabe9 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part d: Simplify nsView::LoadWidget and return early if it fails. r=roc | | | o 7b17bcefb174 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part c: Initialize default widget init data earlier so that it's always available. r=roc | | | o c5945b6a97ed - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part b: Remove nsIDeviceContext::SupportsNativeWidgets because it's not used meaningfully, and will be confusing in content processes. sr=roc | | | o 5452db293694 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part a: Add nsIView::Impl() and nsView::CreateWidget() to get rid of |static_cast<nsview*>(this)|. r=roc | | | o 9407827b5166 - Chris Jones <jones.chris.g@gmail.com> - Bug 582075, part 0.5: Add support for aInitData=NULL to the Windows nsWindow implementation. r=dougt | | | o 88a279b64473 - Chris Jones <jones.chris.g@gmail.com> - Bug 582057, part 0: Log the repaint region bounding rect in DumpPaintEvent. r=roc | | | o cebb111fbfc4 - Robert O'Callahan <roc@ocallahan.org> - Bug 585817. Part 3: Remove nsSVGUtils::GetThebesComputationalSurface and use gfxPlatform::ScreenReferenceSurface instead. r=jwatt | | | o 7b3726c3a580 - Robert O'Callahan <roc@ocallahan.org> - Bug 585817. Part 2: Change nsIPresShell::CreateRenderingContext to GetReferenceRenderingContext, that uses the shared 1x1 surface, and use it all over the place. r=mats,sr=dbaron | | | o b3e968d831ec - Robert O'Callahan <roc@ocallahan.org> - Bug 585817. Part 1: Create a single static 1x1 surface in gfxPlatform that can be used to create contexts for text measurement etc. r=vlad | | | o 55ef0e0529bc - Mounir Lamouri <mounir.lamouri@gmail.com> - Bug 506554 - Implement the CSS3 pseudo-classes :required and :optional for HTML. r=sicking sr=bz a2.0=blocking | | $ hg out comparing with http://hg.mozilla.org/mozilla-central searching for changes 1767c1fb9418 - Chris Jones <jones.chris.g@gmail.com> - Back out bug 585817 and bug 582057 8fccc434c295 - Chris Jones <jones.chris.g@gmail.com> - Merge backout </int></int>
Maintaining a branch of mozilla-central
Let foo be the project you are working on. We assume that the project directory will be http://hg.mozilla.org/projects/foo and is a branch of mozilla-central. We also assume that you want to push some patches on foo and periodically synchronize both repositories.
Modify hgrc
To make things simpler, you can modify the hgrc file in the repository. You can find it here: /path/to/repository/.hg/hgrc
Change it to:
[paths] default = http://hg.mozilla.org/projects/foo m-c = ssh://hg.mozilla.org/mozilla-central default-push = ssh://hg.mozilla.org/projects/foo
Synchronize mozilla-central and foo project
You can push to foo as you would push to any repository but you might want to keep in sync both repositories. To do that, you can run these commands:
hg pull -u # Get all the changes to foo in your repo. hg pull m-c # Get all the changes to mozilla-central in your repo. hg merge # Here, the things might be complex and would need extra carefulness. # resolve merge conflicts, if any hg commit -m "Merge foo with mozilla-central." hg push # Push the merge to foo. hg push m-c # Push the changes to mozilla-central.
Help
Help, I can't push!
If you're trying to push, and you can't, first try this:
$ ssh hg.mozilla.org
If you see output like:
Permission denied (publickey,gssapi-with-mic).
it may have the following reasons:
- you need to configure your username correctly
- you don't have the correct private key in ~/.ssh
- you don't have an hg.mozilla.org account. (If you should have one, file a bug in mozilla.org:Server Operations.)
If you see output like:
You are not allowed to use this system, go away!
then you need to file a bug in mozilla.org:Server Operations.
You should expect the ssh command to disconnect you immediately, since you're not supposed to have shell access. It may give the output:
Could not chdir to home directory : No such file or directory
which is harmless (although some people see it and some people don't).
If you see output like:
ssl required
then you need to configure your default-push correctly.
Branch merges
As part of the rapid release process, the contents of e.g. mozilla-aurora and mozilla-beta are subject to merge events that happen according to the planned branch migration dates. After a branch merge event occurred, if you attempt to use "hg update" in your local tree, you may get error messages such as "abort: crosses branches". In order to fix it, first make sure you don't have local changes, then run "hg update -C -r default".
What's the difference between hg pull and hg update?
hg pull
copies changesets from one repository to another. It only brings changes into your local repository, not your working copy. It will touch the network if you're pulling from a remote repository.
hg update
updates your working copy. It only modifies your working copy. It will not touch the network (unless your directory is on a network filesystem).
How does Mercurial handle line endings?
The Windows version of Mercurial does not automatically convert line endings between Windows and Unix styles. All our repositories use Unix line endings. We need a story for Windows, but right now I have no ideas.
(How about a pre-commit hook that rejects pushes containing CR with a suitably informative error message? We possibly want to make exceptions for certain types of files (at least binary files of course), but we can tweak the hook as necessary as these crop up. Mercurial 1.0 adds a standard hook for this in the win32text extension that we could use/adapt. --jwatt)