bos@104: \chapter{Advanced uses of Mercurial Queues}
bos@104:
bos@104: While it's easy to pick up straightforward uses of Mercurial Queues,
bos@104: use of a little discipline and some of MQ's less frequently used
bos@104: capabilities makes it possible to work in complicated development
bos@104: environments.
bos@104:
bos@105: In this chapter, I will use as an example a technique I have used to
bos@105: manage the development of an Infiniband device driver for the Linux
bos@105: kernel. The driver in question is large (at least as drivers go),
bos@105: with 25,000 lines of code spread across 35 source files. It is
bos@105: maintained by a small team of developers.
bos@104:
bos@104: While much of the material in this chapter is specific to Linux, the
bos@104: same principles apply to any code base for which you're not the
bos@104: primary owner, and upon which you need to do a lot of development.
bos@104:
bos@104: \section{The problem of many targets}
bos@104:
bos@104: The Linux kernel changes rapidly, and has never been internally
bos@104: stable; developers frequently make drastic changes between releases.
bos@104: This means that a version of the driver that works well with a
bos@104: particular released version of the kernel will not even \emph{compile}
bos@104: correctly against, typically, any other version.
bos@104:
bos@104: To maintain a driver, we have to keep a number of distinct versions of
bos@104: Linux in mind.
bos@104: \begin{itemize}
bos@104: \item One target is the main Linux kernel development tree.
bos@104: Maintenance of the code is in this case partly shared by other
bos@104: developers in the kernel community, who make ``drive-by''
bos@104: modifications to the driver as they develop and refine kernel
bos@104: subsystems.
bos@104: \item We also maintain a number of ``backports'' to older versions of
bos@104: the Linux kernel, to support the needs of customers who are running
bos@105: older Linux distributions that do not incorporate our drivers. (To
bos@105: \emph{backport} a piece of code is to modify it to work in an older
bos@105: version of its target environment than the version it was developed
bos@105: for.)
bos@104: \item Finally, we make software releases on a schedule that is
bos@104: necessarily not aligned with those used by Linux distributors and
bos@104: kernel developers, so that we can deliver new features to customers
bos@104: without forcing them to upgrade their entire kernels or
bos@104: distributions.
bos@104: \end{itemize}
bos@104:
bos@104: \subsection{Tempting approaches that don't work well}
bos@104:
bos@104: There are two ``standard'' ways to maintain a piece of software that
bos@104: has to target many different environments.
bos@104:
bos@104: The first is to maintain a number of branches, each intended for a
bos@104: single target. The trouble with this approach is that you must
bos@104: maintain iron discipline in the flow of changes between repositories.
bos@104: A new feature or bug fix must start life in a ``pristine'' repository,
bos@104: then percolate out to every backport repository. Backport changes are
bos@104: more limited in the branches they should propagate to; a backport
bos@104: change that is applied to a branch where it doesn't belong will
bos@104: probably stop the driver from compiling.
bos@104:
bos@104: The second is to maintain a single source tree filled with conditional
bos@104: statements that turn chunks of code on or off depending on the
bos@104: intended target. Because these ``ifdefs'' are not allowed in the
bos@104: Linux kernel tree, a manual or automatic process must be followed to
bos@104: strip them out and yield a clean tree. A code base maintained in this
bos@104: fashion rapidly becomes a rat's nest of conditional blocks that are
bos@104: difficult to understand and maintain.
bos@104:
bos@104: Neither of these approaches is well suited to a situation where you
bos@104: don't ``own'' the canonical copy of a source tree. In the case of a
bos@104: Linux driver that is distributed with the standard kernel, Linus's
bos@104: tree contains the copy of the code that will be treated by the world
bos@104: as canonical. The upstream version of ``my'' driver can be modified
bos@104: by people I don't know, without me even finding out about it until
bos@104: after the changes show up in Linus's tree.
bos@104:
bos@104: These approaches have the added weakness of making it difficult to
bos@104: generate well-formed patches to submit upstream.
bos@104:
bos@104: In principle, Mercurial Queues seems like a good candidate to manage a
bos@104: development scenario such as the above. While this is indeed the
bos@104: case, MQ contains a few added features that make the job more
bos@104: pleasant.
bos@104:
bos@105: \section{Conditionally applying patches with
bos@105: guards}
bos@104:
bos@104: Perhaps the best way to maintain sanity with so many targets is to be
bos@104: able to choose specific patches to apply for a given situation. MQ
bos@104: provides a feature called ``guards'' (which originates with quilt's
bos@104: \texttt{guards} command) that does just this. To start off, let's
bos@104: create a simple repository for experimenting in.
bos@104: \interaction{mq.guards.init}
bos@104: This gives us a tiny repository that contains two patches that don't
bos@104: have any dependencies on each other, because they touch different files.
bos@104:
bos@104: The idea behind conditional application is that you can ``tag'' a
bos@104: patch with a \emph{guard}, which is simply a text string of your
bos@104: choosing, then tell MQ to select specific guards to use when applying
bos@104: patches. MQ will then either apply, or skip over, a guarded patch,
bos@104: depending on the guards that you have selected.
bos@104:
bos@104: A patch can have an arbitrary number of guards;
bos@104: each one is \emph{positive} (``apply this patch if this guard is
bos@104: selected'') or \emph{negative} (``skip this patch if this guard is
bos@104: selected''). A patch with no guards is always applied.
bos@104:
bos@104: \section{Controlling the guards on a patch}
bos@104:
bos@104: The \hgcmd{qguard} command lets you determine which guards should
bos@104: apply to a patch, or display the guards that are already in effect.
bos@104: Without any arguments, it displays the guards on the current topmost
bos@104: patch.
bos@104: \interaction{mq.guards.qguard}
bos@104: To set a positive guard on a patch, prefix the name of the guard with
bos@104: a ``\texttt{+}''.
bos@104: \interaction{mq.guards.qguard.pos}
bos@104: To set a negative guard on a patch, prefix the name of the guard with
bos@104: a ``\texttt{-}''.
bos@104: \interaction{mq.guards.qguard.neg}
bos@104:
bos@104: \begin{note}
bos@104: The \hgcmd{qguard} command \emph{sets} the guards on a patch; it
bos@104: doesn't \emph{modify} them. What this means is that if you run
bos@104: \hgcmdargs{qguard}{+a +b} on a patch, then \hgcmdargs{qguard}{+c} on
bos@104: the same patch, the \emph{only} guard that will be set on it
bos@104: afterwards is \texttt{+c}.
bos@104: \end{note}
bos@104:
bos@104: Mercurial stores guards in the \sfilename{series} file; the form in
bos@104: which they are stored is easy both to understand and to edit by hand.
bos@104: (In other words, you don't have to use the \hgcmd{qguard} command if
bos@104: you don't want to; it's okay to simply edit the \sfilename{series}
bos@104: file.)
bos@104: \interaction{mq.guards.series}
bos@104:
bos@104: \section{Selecting the guards to use}
bos@104:
bos@104: The \hgcmd{qselect} command determines which guards are active at a
bos@104: given time. The effect of this is to determine which patches MQ will
bos@104: apply the next time you run \hgcmd{qpush}. It has no other effect; in
bos@104: particular, it doesn't do anything to patches that are already
bos@104: applied.
bos@104:
bos@104: With no arguments, the \hgcmd{qselect} command lists the guards
bos@104: currently in effect, one per line of output. Each argument is treated
bos@104: as the name of a guard to apply.
bos@104: \interaction{mq.guards.qselect.foo}
bos@104: In case you're interested, the currently selected guards are stored in
bos@104: the \sfilename{guards} file.
bos@104: \interaction{mq.guards.qselect.cat}
bos@104: We can see the effect the selected guards have when we run
bos@104: \hgcmd{qpush}.
bos@104: \interaction{mq.guards.qselect.qpush}
bos@104:
bos@104: A guard cannot start with a ``\texttt{+}'' or ``\texttt{-}''
bos@106: character. The name of a guard must not contain white space, but most
bos@106: othter characters are acceptable. If you try to use a guard with an
bos@106: invalid name, MQ will complain:
bos@106: \interaction{mq.guards.qselect.error}
bos@104: Changing the selected guards changes the patches that are applied.
bos@106: \interaction{mq.guards.qselect.quux}
bos@105: You can see in the example below that negative guards take precedence
bos@105: over positive guards.
bos@104: \interaction{mq.guards.qselect.foobar}
bos@104:
bos@105: \section{MQ's rules for applying patches}
bos@105:
bos@105: The rules that MQ uses when deciding whether to apply a patch
bos@105: are as follows.
bos@105: \begin{itemize}
bos@105: \item A patch that has no guards is always applied.
bos@105: \item If the patch has any negative guard that matches any currently
bos@105: selected guard, the patch is skipped.
bos@105: \item If the patch has any positive guard that matches any currently
bos@105: selected guard, the patch is applied.
bos@105: \item If the patch has positive or negative guards, but none matches
bos@105: any currently selected guard, the patch is skipped.
bos@105: \end{itemize}
bos@105:
bos@105: \section{Trimming the work environment}
bos@105:
bos@105: In working on the device driver I mentioned earlier, I don't apply the
bos@105: patches to a normal Linux kernel tree. Instead, I use a repository
bos@105: that contains only a snapshot of the source files and headers that are
bos@105: relevant to Infiniband development. This repository is~1\% the size
bos@105: of a kernel repository, so it's easier to work with.
bos@105:
bos@105: I then choose a ``base'' version on top of which the patches are
bos@105: applied. This is a snapshot of the Linux kernel tree as of a revision
bos@105: of my choosing. When I take the snapshot, I record the changeset ID
bos@105: from the kernel repository in the commit message. Since the snapshot
bos@105: preserves the ``shape'' and content of the relevant parts of the
bos@105: kernel tree, I can apply my patches on top of either my tiny
bos@105: repository or a normal kernel tree.
bos@105:
bos@105: Normally, the base tree atop which the patches apply should be a
bos@105: snapshot of a very recent upstream tree. This best facilitates the
bos@105: development of patches that can easily be submitted upstream with few
bos@105: or no modifications.
bos@105:
bos@105: \section{Dividing up the \sfilename{series} file}
bos@105:
bos@105: I categorise the patches in the \sfilename{series} file into a number
bos@105: of logical groups. Each section of like patches begins with a block
bos@105: of comments that describes the purpose of the patches that follow.
bos@105:
bos@105: The sequence of patch groups that I maintain follows. The ordering of
bos@105: these groups is important; I'll describe why after I introduce the
bos@105: groups.
bos@105: \begin{itemize}
bos@105: \item The ``accepted'' group. Patches that the development team has
bos@105: submitted to the maintainer of the Infiniband subsystem, and which
bos@105: he has accepted, but which are not present in the snapshot that the
bos@105: tiny repository is based on. These are ``read only'' patches,
bos@105: present only to transform the tree into a similar state as it is in
bos@105: the upstream maintainer's repository.
bos@105: \item The ``rework'' group. Patches that I have submitted, but that
bos@105: the upstream maintainer has requested modifications to before he
bos@105: will accept them.
bos@105: \item The ``pending'' group. Patches that I have not yet submitted to
bos@105: the upstream maintainer, but which we have finished working on.
bos@105: These will be ``read only'' for a while. If the upstream maintainer
bos@105: accepts them upon submission, I'll move them to the end of the
bos@105: ``accepted'' group. If he requests that I modify any, I'll move
bos@105: them to the beginning of the ``rework'' group.
bos@105: \item The ``in progress'' group. Patches that are actively being
bos@105: developed, and should not be submitted anywhere yet.
bos@105: \item The ``backport'' group. Patches that adapt the source tree to
bos@105: older versions of the kernel tree.
bos@105: \item The ``do not ship'' group. Patches that for some reason should
bos@105: never be submitted upstream. For example, one such patch might
bos@105: change embedded driver identification strings to make it easier to
bos@105: distinguish, in the field, between an out-of-tree version of the
bos@105: driver and a version shipped by a distribution vendor.
bos@105: \end{itemize}
bos@105:
bos@105: Now to return to the reasons for ordering groups of patches in this
bos@105: way. We would like the lowest patches in the stack to be as stable as
bos@105: possible, so that we will not need to rework higher patches due to
bos@105: changes in context. Putting patches that will never be changed first
bos@105: in the \sfilename{series} file serves this purpose.
bos@105:
bos@105: We would also like the patches that we know we'll need to modify to be
bos@105: applied on top of a source tree that resembles the upstream tree as
bos@105: closely as possible. This is why we keep accepted patches around for
bos@105: a while.
bos@105:
bos@105: The ``backport'' and ``do not ship'' patches float at the end of the
bos@106: \sfilename{series} file. The backport patches must be applied on top
bos@106: of all other patches, and the ``do not ship'' patches might as well
bos@106: stay out of harm's way.
bos@106:
bos@106: \section{Maintaining the patch series}
bos@106:
bos@106: In my work, I use a number of guards to control which patches are to
bos@106: be applied.
bos@106:
bos@106: \begin{itemize}
bos@106: \item ``Accepted'' patches are guarded with \texttt{accepted}. I
bos@106: enable this guard most of the time. When I'm applying the patches
bos@106: on top of a tree where the patches are already present, I can turn
bos@106: this patch off, and the paptches that follow it will apply cleanly.
bos@106: \item Patches that are ``finished'', but not yet submitted, have no
bos@106: guards. If I'm applying the patch stack to a copy of the upstream
bos@106: tree, I don't need to enable any guards in order to get a reasonably
bos@106: safe source tree.
bos@106: \item Those patches that need reworking before being resubmitted are
bos@106: guarded with \texttt{rework}.
bos@106: \item For those patches that are still under development, I use
bos@106: \texttt{devel}.
bos@106: \item A backport patch may have several guards, one for each version
bos@106: of the kernel to which it applies. For example, a patch that
bos@106: backports a piece of code to~2.6.9 will have a~\texttt{2.6.9} guard.
bos@106: \end{itemize}
bos@106: This variety of guards gives me considerable flexibility in
bos@106: qdetermining what kind of source tree I want to end up with. For most
bos@106: situations, the selection of appropriate guards is automated during
bos@106: the build process, but I can manually tune the guards to use for less
bos@106: common circumstances.
bos@106:
bos@106: \subsection{The art of writing backport patches}
bos@106:
bos@106: Using MQ, writing a backport patch is a simple process. All such a
bos@106: patch has to do is modify a piece of code that uses a kernel feature
bos@106: not present in the older version of the kernel, so that the driver
bos@106: continues to work correctly under that older version.
bos@106:
bos@106: A useful goal when writing a good backport patch is to make your code
bos@106: look as if it was written for the older version of the kernel you're
bos@106: targeting. The less obtrusive the patch, the easier it will be to
bos@106: understand and maintain. If you're writing a collection of backport
bos@106: patches to avoid the ``rat's nest'' effect of lots of
bos@106: \texttt{\#ifdef}s (hunks of source code that are only used
bos@106: conditionally) in your code, don't introduce version-dependent
bos@106: \texttt{\#ifdef}s into the patches. Instead, write several patches,
bos@106: each of which makes unconditional changes, and control their
bos@106: application using guards.
bos@106:
bos@106: There are two reasons to divide backport patches into a distinct
bos@106: group, away from the ``regular'' patches whose effects they modify.
bos@106: The first is that intermingling the two makes it more difficult to use
bos@106: a tool like the \hgext{patchbomb} extension to automate the process of
bos@106: submitting the patches to an upstream maintainer. The second is that
bos@106: a backport patch could perturb the context in which a subsequent
bos@106: regular patch is applied, making it impossible to apply the regular
bos@106: patch cleanly \emph{without} the earlier backport patch already being
bos@106: applied.
bos@106:
bos@106: \section{Useful tips for developing with MQ}
bos@106:
bos@106: \subsection{Organising patches in directories}
bos@106:
bos@106: If you're working on a substantial project with MQ, it's not difficult
bos@106: to accumulate a large number of patches. For example, I have one
bos@106: patch repository that contains over 250 patches.
bos@106:
bos@106: If you can group these patches into separate logical categories, you
bos@106: can if you like store them in different directories; MQ has no
bos@106: problems with patch names that contain path separators.
bos@106:
bos@106: \subsection{Viewing the history of a patch}
bos@106: \label{mq-collab:tips:interdiff}
bos@106:
bos@106: If you're developing a set of patches over a long time, it's a good
bos@106: idea to maintain them in a repository, as discussed in
bos@106: section~\ref{sec:mq:repo}. If you do so, you'll quickly discover that
bos@106: using the \hgcmd{diff} command to look at the history of changes to a
bos@106: patch is unworkable. This is in part because you're looking at the
bos@106: second derivative of the real code (a diff of a diff), but also
bos@106: because MQ adds noise to the process by modifying time stamps and
bos@106: directory names when it updates a patch.
bos@106:
bos@106: However, you can use the \hgext{extdiff} extension, which is bundled
bos@106: with Mercurial, to turn a diff of two versions of a patch into
bos@106: something readable. To do this, you will need a third-party package
bos@106: called \package{patchutils}~\cite{web:patchutils}. This provides a
bos@106: command named \command{interdiff}, which shows the differences between
bos@106: two diffs as a diff. Used on two versions of the same diff, it
bos@106: generates a diff that represents the diff from the first to the second
bos@106: version.
bos@106:
bos@106: You can enable the \hgext{extdiff} extension in the usual way, by
bos@106: adding a line to the \rcsection{extensions} section of your \hgrc.
bos@106: \begin{codesample2}
bos@106: [extensions]
bos@106: extdiff =
bos@106: \end{codesample2}
bos@106: The \command{interdiff} command expects to be passed the names of two
bos@106: files, but the \hgext{extdiff} extension passes the program it runs a
bos@106: pair of directories, each of which can contain an arbitrary number of
bos@106: files. We thus need a small program that will run \command{interdiff}
bos@106: on each pair of files in these two directories. This program is
bos@106: available as \sfilename{hg-interdiff} in the \dirname{examples}
bos@106: directory of the source code repository that accompanies this book.
bos@106: \excode{hg-interdiff}
bos@106:
bos@106: With the \sfilename{hg-interdiff} program in your shell's search path,
bos@106: you can run it as follows, from inside an MQ patch directory:
bos@106: \begin{codesample2}
bos@106: hg extdiff -p hg-interdiff -r A:B my-change.patch
bos@106: \end{codesample2}
bos@106: Since you'll probably want to use this long-winded command a lot, you
bos@106: can get \hgext{hgext} to make it available as a normal Mercurial
bos@106: command, again by editing your \hgrc.
bos@106: \begin{codesample2}
bos@106: [extdiff]
bos@106: cmd.interdiff = hg-interdiff
bos@106: \end{codesample2}
bos@106: This directs \hgext{hgext} to make an \texttt{interdiff} command
bos@106: available, so you can now shorten the previous invocation of
bos@106: \hgcmd{extdiff} to something a little more wieldy.
bos@106: \begin{codesample2}
bos@106: hg interdiff -r A:B my-change.patch
bos@106: \end{codesample2}
bos@105:
bos@107: \begin{note}
bos@107: The \command{interdiff} command works well only if the underlying
bos@107: files against which versions of a patch are generated remain the
bos@107: same. If you create a patch, modify the underlying files, and then
bos@107: regenerate the patch, \command{interdiff} may not produce useful
bos@107: output.
bos@107: \end{note}
bos@107:
bos@104: %%% Local Variables:
bos@104: %%% mode: latex
bos@104: %%% TeX-master: "00book"
bos@104: %%% End: