igor@374: \chapter{Encontrar y arreglar sus equivocaciones}
jerojasro@343: \label{chap:undo}
jerojasro@343: 
igor@374: Errar es humano, pero tratar adecuadamente las consecuencias requiere
igor@374: un sistema de control de revisiones de primera categoría.  En este
igor@374: capítulo, discutiremos algunas técnicas que puede usar cuando
igor@374: encuentra que hay un problema enraizado en su proyecto.  Mercurial
igor@374: tiene unas características poderosas que le ayudarán a isolar las
igor@374: fuentes de los problemas, y a dar cuenta de ellas apropiadamente.
igor@374: 
igor@374: \section{Borrar la historia local}
igor@374: 
igor@374: \subsection{La consignación accidental}
igor@374: 
igor@374: Tengo el problema ocasional, pero persistente de teclear más rápido de
igor@374: lo que pienso, que aveces resulta en consignar un conjunto de cambios
igor@374: incompleto o simplemente malo. En mi caso, el conjunto de cambios
igor@374: incompleto consiste en que creé un nuevo archivo fuente, pero olvidé
igor@374: hacerle \hgcmd{add}.  Un conjunto de cambios``simplemente malo'' no es
igor@374: tan común, pero sí resulta muy molesto.
igor@374: 
igor@374: \subsection{Retroceder una transacción}
jerojasro@343: \label{sec:undo:rollback}
jerojasro@343: 
igor@374: En la sección~\ref{sec:concepts:txn}, mencioné que Mercurial trata
igor@374: modificación a un repositorio como una \emph{transacción}.  Cada vez
igor@374: que consigna un conjunto de cambios o lo jala de otro repositorio,
igor@374: Mercurial recuerda lo que hizo.  Puede deshacer, o \emph{retroceder},
igor@374: exactamente una de tales acciones usando la orden \hgcmd{rollback}.
igor@374: (Ver en la sección~\ref{sec:undo:rollback-after-push} una anotación
igor@374: importante acerca del uso de esta orden.)
igor@374: 
igor@374: A continuación una equivocación que me sucede frecuentemente:
igor@374: consignar un cambio en el cual he creado un nuevo fichero, pero he
igor@374: olvidado hacerle \hgcmd{add}.
jerojasro@343: \interaction{rollback.commit}
igor@374: La salida de \hgcmd{status} después de la consignación confirma
igor@374: inmediatamente este error.
jerojasro@343: \interaction{rollback.status}
igor@374: La consignación capturó los cambios en el archivo \filename{a}, pero
igor@374: no el nuevo fichero \filename{b}.  Si yo publicara este conjunto de
igor@374: cambios a un repositorio compartido con un colega, es bastante
igor@374: probable que algo en \filename{a} se refiriera a \filename{b}, el cual
igor@374: podría no estar presente cuando jalen mis cambios del repositorio.  Me
igor@374: convertiría el sujeto de cierta indignación.
igor@374: 
igor@374: Como sea, la suerte me acompaña---Encontré mi error antes de publicar
igor@374: el conjunto de cambios.  Uso la orden \hgcmd{rollback}, y Mercurial
igor@374: hace desaparecer el último conjunto de cambios.
jerojasro@343: \interaction{rollback.rollback}
igor@374: El conjunto de cambios ya no está en la historia del repositorio, y el
igor@374: directorio de trabajo cree que el fichero \filename{a} ha sido
igor@374: modificado.  La consignación y el retroceso dejaron el directorio de
igor@374: trabajo exactamente como estaban antes de la consignación; el conjunto
igor@374: de cambios ha sido eliminado totlamente. Ahora puedo hacer \hgcmd{add}
igor@374: al fichero \filename{b}, y hacer de nuevo la consignación.
jerojasro@343: \interaction{rollback.add}
jerojasro@343: 
igor@374: \subsection{Erroneamente jalado}
igor@374: 
igor@374: Mantener ramas de desarrollo separadas de un proyecto en distintos
igor@374: repositorios es una práctica común con Mercurial.  Su equipo de
igor@374: desarrollo puede tener un repositorio compartido para la versión ``0.9''
igor@374: y otra con cambios distintos para la versión ``1.0''.
igor@374: 
igor@374: Con este escenario, puede imaginar las consecuencias si tuviera un
igor@374: repositorio local ``0.9'', y jalara accidentalmente los cambios del
igor@374: repositorio compartido de la versión ``1.0'' en este.  En el peor de
igor@374: los casos, por falta de atención, es posible que publique tales
igor@374: cambios en el árbol compartido ``0.9'', confundiendo a todo su equipo
igor@374: de trabajo(pero no se preocupe, volveremos a este terrorífico
igor@374: escenario posteriormente).  En todo caso, es muy probable que usted se
igor@374: de cuenta inmediatamente, dado que Mercurial mostrará el URL de donde
igor@374: está jalando, o que vea jalando una sospechosa gran cantidad de
igor@374: cambios en el repositorio.
igor@374: 
igor@374: La orden \hgcmd{rollback} command will work nicely to expunge all of the
jerojasro@343: changesets that you just pulled.  Mercurial groups all changes from
jerojasro@343: one \hgcmd{pull} into a single transaction, so one \hgcmd{rollback} is
jerojasro@343: all you need to undo this mistake.
jerojasro@343: 
jerojasro@343: \subsection{Rolling back is useless once you've pushed}
jerojasro@343: \label{sec:undo:rollback-after-push}
jerojasro@343: 
jerojasro@343: The value of the \hgcmd{rollback} command drops to zero once you've
jerojasro@343: pushed your changes to another repository.  Rolling back a change
jerojasro@343: makes it disappear entirely, but \emph{only} in the repository in
jerojasro@343: which you perform the \hgcmd{rollback}.  Because a rollback eliminates
jerojasro@343: history, there's no way for the disappearance of a change to propagate
jerojasro@343: between repositories.
jerojasro@343: 
jerojasro@343: If you've pushed a change to another repository---particularly if it's
jerojasro@343: a shared repository---it has essentially ``escaped into the wild,''
jerojasro@343: and you'll have to recover from your mistake in a different way.  What
jerojasro@343: will happen if you push a changeset somewhere, then roll it back, then
jerojasro@343: pull from the repository you pushed to, is that the changeset will
jerojasro@343: reappear in your repository.
jerojasro@343: 
jerojasro@343: (If you absolutely know for sure that the change you want to roll back
jerojasro@343: is the most recent change in the repository that you pushed to,
jerojasro@343: \emph{and} you know that nobody else could have pulled it from that
jerojasro@343: repository, you can roll back the changeset there, too, but you really
jerojasro@343: should really not rely on this working reliably.  If you do this,
jerojasro@343: sooner or later a change really will make it into a repository that
jerojasro@343: you don't directly control (or have forgotten about), and come back to
jerojasro@343: bite you.)
jerojasro@343: 
jerojasro@343: \subsection{You can only roll back once}
jerojasro@343: 
jerojasro@343: Mercurial stores exactly one transaction in its transaction log; that
jerojasro@343: transaction is the most recent one that occurred in the repository.
jerojasro@343: This means that you can only roll back one transaction.  If you expect
jerojasro@343: to be able to roll back one transaction, then its predecessor, this is
jerojasro@343: not the behaviour you will get.
jerojasro@343: \interaction{rollback.twice}
jerojasro@343: Once you've rolled back one transaction in a repository, you can't
jerojasro@343: roll back again in that repository until you perform another commit or
jerojasro@343: pull.
jerojasro@343: 
jerojasro@343: \section{Reverting the mistaken change}
jerojasro@343: 
jerojasro@343: If you make a modification to a file, and decide that you really
jerojasro@343: didn't want to change the file at all, and you haven't yet committed
jerojasro@343: your changes, the \hgcmd{revert} command is the one you'll need.  It
jerojasro@343: looks at the changeset that's the parent of the working directory, and
jerojasro@343: restores the contents of the file to their state as of that changeset.
jerojasro@343: (That's a long-winded way of saying that, in the normal case, it
jerojasro@343: undoes your modifications.)
jerojasro@343: 
jerojasro@343: Let's illustrate how the \hgcmd{revert} command works with yet another
jerojasro@343: small example.  We'll begin by modifying a file that Mercurial is
jerojasro@343: already tracking.
jerojasro@343: \interaction{daily.revert.modify}
jerojasro@343: If we don't want that change, we can simply \hgcmd{revert} the file.
jerojasro@343: \interaction{daily.revert.unmodify}
jerojasro@343: The \hgcmd{revert} command provides us with an extra degree of safety
jerojasro@343: by saving our modified file with a \filename{.orig} extension.
jerojasro@343: \interaction{daily.revert.status}
jerojasro@343: 
jerojasro@343: Here is a summary of the cases that the \hgcmd{revert} command can
jerojasro@343: deal with.  We will describe each of these in more detail in the
jerojasro@343: section that follows.
jerojasro@343: \begin{itemize}
jerojasro@343: \item If you modify a file, it will restore the file to its unmodified
jerojasro@343:   state.
jerojasro@343: \item If you \hgcmd{add} a file, it will undo the ``added'' state of
jerojasro@343:   the file, but leave the file itself untouched.
jerojasro@343: \item If you delete a file without telling Mercurial, it will restore
jerojasro@343:   the file to its unmodified contents.
jerojasro@343: \item If you use the \hgcmd{remove} command to remove a file, it will
jerojasro@343:   undo the ``removed'' state of the file, and restore the file to its
jerojasro@343:   unmodified contents.
jerojasro@343: \end{itemize}
jerojasro@343: 
jerojasro@343: \subsection{File management errors}
jerojasro@343: \label{sec:undo:mgmt}
jerojasro@343: 
jerojasro@343: The \hgcmd{revert} command is useful for more than just modified
jerojasro@343: files.  It lets you reverse the results of all of Mercurial's file
jerojasro@343: management commands---\hgcmd{add}, \hgcmd{remove}, and so on.
jerojasro@343: 
jerojasro@343: If you \hgcmd{add} a file, then decide that in fact you don't want
jerojasro@343: Mercurial to track it, use \hgcmd{revert} to undo the add.  Don't
jerojasro@343: worry; Mercurial will not modify the file in any way.  It will just
jerojasro@343: ``unmark'' the file.
jerojasro@343: \interaction{daily.revert.add}
jerojasro@343: 
jerojasro@343: Similarly, if you ask Mercurial to \hgcmd{remove} a file, you can use
jerojasro@343: \hgcmd{revert} to restore it to the contents it had as of the parent
jerojasro@343: of the working directory.
jerojasro@343: \interaction{daily.revert.remove}
jerojasro@343: This works just as well for a file that you deleted by hand, without
jerojasro@343: telling Mercurial (recall that in Mercurial terminology, this kind of
jerojasro@343: file is called ``missing'').
jerojasro@343: \interaction{daily.revert.missing}
jerojasro@343: 
jerojasro@343: If you revert a \hgcmd{copy}, the copied-to file remains in your
jerojasro@343: working directory afterwards, untracked.  Since a copy doesn't affect
jerojasro@343: the copied-from file in any way, Mercurial doesn't do anything with
jerojasro@343: the copied-from file.
jerojasro@343: \interaction{daily.revert.copy}
jerojasro@343: 
jerojasro@343: \subsubsection{A slightly special case: reverting a rename}
jerojasro@343: 
jerojasro@343: If you \hgcmd{rename} a file, there is one small detail that
jerojasro@343: you should remember.  When you \hgcmd{revert} a rename, it's not
jerojasro@343: enough to provide the name of the renamed-to file, as you can see
jerojasro@343: here.
jerojasro@343: \interaction{daily.revert.rename}
jerojasro@343: As you can see from the output of \hgcmd{status}, the renamed-to file
jerojasro@343: is no longer identified as added, but the renamed-\emph{from} file is
jerojasro@343: still removed!  This is counter-intuitive (at least to me), but at
jerojasro@343: least it's easy to deal with.
jerojasro@343: \interaction{daily.revert.rename-orig}
jerojasro@343: So remember, to revert a \hgcmd{rename}, you must provide \emph{both}
jerojasro@343: the source and destination names.  
jerojasro@343: 
jerojasro@343: % TODO: the output doesn't look like it will be removed!
jerojasro@343: 
jerojasro@343: (By the way, if you rename a file, then modify the renamed-to file,
jerojasro@343: then revert both components of the rename, when Mercurial restores the
jerojasro@343: file that was removed as part of the rename, it will be unmodified.
jerojasro@343: If you need the modifications in the renamed-to file to show up in the
jerojasro@343: renamed-from file, don't forget to copy them over.)
jerojasro@343: 
jerojasro@343: These fiddly aspects of reverting a rename arguably constitute a small
jerojasro@343: bug in Mercurial.
jerojasro@343: 
jerojasro@343: \section{Dealing with committed changes}
jerojasro@343: 
jerojasro@343: Consider a case where you have committed a change $a$, and another
jerojasro@343: change $b$ on top of it; you then realise that change $a$ was
jerojasro@343: incorrect.  Mercurial lets you ``back out'' an entire changeset
jerojasro@343: automatically, and building blocks that let you reverse part of a
jerojasro@343: changeset by hand.
jerojasro@343: 
jerojasro@343: Before you read this section, here's something to keep in mind: the
jerojasro@343: \hgcmd{backout} command undoes changes by \emph{adding} history, not
jerojasro@343: by modifying or erasing it.  It's the right tool to use if you're
jerojasro@343: fixing bugs, but not if you're trying to undo some change that has
jerojasro@343: catastrophic consequences.  To deal with those, see
jerojasro@343: section~\ref{sec:undo:aaaiiieee}.
jerojasro@343: 
jerojasro@343: \subsection{Backing out a changeset}
jerojasro@343: 
jerojasro@343: The \hgcmd{backout} command lets you ``undo'' the effects of an entire
jerojasro@343: changeset in an automated fashion.  Because Mercurial's history is
jerojasro@343: immutable, this command \emph{does not} get rid of the changeset you
jerojasro@343: want to undo.  Instead, it creates a new changeset that
jerojasro@343: \emph{reverses} the effect of the to-be-undone changeset.
jerojasro@343: 
jerojasro@343: The operation of the \hgcmd{backout} command is a little intricate, so
jerojasro@343: let's illustrate it with some examples.  First, we'll create a
jerojasro@343: repository with some simple changes.
jerojasro@343: \interaction{backout.init}
jerojasro@343: 
jerojasro@343: The \hgcmd{backout} command takes a single changeset ID as its
jerojasro@343: argument; this is the changeset to back out.  Normally,
jerojasro@343: \hgcmd{backout} will drop you into a text editor to write a commit
jerojasro@343: message, so you can record why you're backing the change out.  In this
jerojasro@343: example, we provide a commit message on the command line using the
jerojasro@343: \hgopt{backout}{-m} option.
jerojasro@343: 
jerojasro@343: \subsection{Backing out the tip changeset}
jerojasro@343: 
jerojasro@343: We're going to start by backing out the last changeset we committed.
jerojasro@343: \interaction{backout.simple}
jerojasro@343: You can see that the second line from \filename{myfile} is no longer
jerojasro@343: present.  Taking a look at the output of \hgcmd{log} gives us an idea
jerojasro@343: of what the \hgcmd{backout} command has done.
jerojasro@343: \interaction{backout.simple.log}
jerojasro@343: Notice that the new changeset that \hgcmd{backout} has created is a
jerojasro@343: child of the changeset we backed out.  It's easier to see this in
jerojasro@343: figure~\ref{fig:undo:backout}, which presents a graphical view of the
jerojasro@343: change history.  As you can see, the history is nice and linear.
jerojasro@343: 
jerojasro@343: \begin{figure}[htb]
jerojasro@343:   \centering
jerojasro@343:   \grafix{undo-simple}
jerojasro@343:   \caption{Backing out a change using the \hgcmd{backout} command}
jerojasro@343:   \label{fig:undo:backout}
jerojasro@343: \end{figure}
jerojasro@343: 
jerojasro@343: \subsection{Backing out a non-tip change}
jerojasro@343: 
jerojasro@343: If you want to back out a change other than the last one you
jerojasro@343: committed, pass the \hgopt{backout}{--merge} option to the
jerojasro@343: \hgcmd{backout} command.
jerojasro@343: \interaction{backout.non-tip.clone}
jerojasro@343: This makes backing out any changeset a ``one-shot'' operation that's
jerojasro@343: usually simple and fast.
jerojasro@343: \interaction{backout.non-tip.backout}
jerojasro@343: 
jerojasro@343: If you take a look at the contents of \filename{myfile} after the
jerojasro@343: backout finishes, you'll see that the first and third changes are
jerojasro@343: present, but not the second.
jerojasro@343: \interaction{backout.non-tip.cat}
jerojasro@343: 
jerojasro@343: As the graphical history in figure~\ref{fig:undo:backout-non-tip}
jerojasro@343: illustrates, Mercurial actually commits \emph{two} changes in this
jerojasro@343: kind of situation (the box-shaped nodes are the ones that Mercurial
jerojasro@343: commits automatically).  Before Mercurial begins the backout process,
jerojasro@343: it first remembers what the current parent of the working directory
jerojasro@343: is.  It then backs out the target changeset, and commits that as a
jerojasro@343: changeset.  Finally, it merges back to the previous parent of the
jerojasro@343: working directory, and commits the result of the merge.
jerojasro@343: 
jerojasro@343: % TODO: to me it looks like mercurial doesn't commit the second merge automatically!
jerojasro@343: 
jerojasro@343: \begin{figure}[htb]
jerojasro@343:   \centering
jerojasro@343:   \grafix{undo-non-tip}
jerojasro@343:   \caption{Automated backout of a non-tip change using the \hgcmd{backout} command}
jerojasro@343:   \label{fig:undo:backout-non-tip}
jerojasro@343: \end{figure}
jerojasro@343: 
jerojasro@343: The result is that you end up ``back where you were'', only with some
jerojasro@343: extra history that undoes the effect of the changeset you wanted to
jerojasro@343: back out.
jerojasro@343: 
jerojasro@343: \subsubsection{Always use the \hgopt{backout}{--merge} option}
jerojasro@343: 
jerojasro@343: In fact, since the \hgopt{backout}{--merge} option will do the ``right
jerojasro@343: thing'' whether or not the changeset you're backing out is the tip
jerojasro@343: (i.e.~it won't try to merge if it's backing out the tip, since there's
jerojasro@343: no need), you should \emph{always} use this option when you run the
jerojasro@343: \hgcmd{backout} command.
jerojasro@343: 
jerojasro@343: \subsection{Gaining more control of the backout process}
jerojasro@343: 
jerojasro@343: While I've recommended that you always use the
jerojasro@343: \hgopt{backout}{--merge} option when backing out a change, the
jerojasro@343: \hgcmd{backout} command lets you decide how to merge a backout
jerojasro@343: changeset.  Taking control of the backout process by hand is something
jerojasro@343: you will rarely need to do, but it can be useful to understand what
jerojasro@343: the \hgcmd{backout} command is doing for you automatically.  To
jerojasro@343: illustrate this, let's clone our first repository, but omit the
jerojasro@343: backout change that it contains.
jerojasro@343: 
jerojasro@343: \interaction{backout.manual.clone}
jerojasro@343: As with our earlier example, We'll commit a third changeset, then back
jerojasro@343: out its parent, and see what happens.
jerojasro@343: \interaction{backout.manual.backout} 
jerojasro@343: Our new changeset is again a descendant of the changeset we backout
jerojasro@343: out; it's thus a new head, \emph{not} a descendant of the changeset
jerojasro@343: that was the tip.  The \hgcmd{backout} command was quite explicit in
jerojasro@343: telling us this.
jerojasro@343: \interaction{backout.manual.log}
jerojasro@343: 
jerojasro@343: Again, it's easier to see what has happened by looking at a graph of
jerojasro@343: the revision history, in figure~\ref{fig:undo:backout-manual}.  This
jerojasro@343: makes it clear that when we use \hgcmd{backout} to back out a change
jerojasro@343: other than the tip, Mercurial adds a new head to the repository (the
jerojasro@343: change it committed is box-shaped).
jerojasro@343: 
jerojasro@343: \begin{figure}[htb]
jerojasro@343:   \centering
jerojasro@343:   \grafix{undo-manual}
jerojasro@343:   \caption{Backing out a change using the \hgcmd{backout} command}
jerojasro@343:   \label{fig:undo:backout-manual}
jerojasro@343: \end{figure}
jerojasro@343: 
jerojasro@343: After the \hgcmd{backout} command has completed, it leaves the new
jerojasro@343: ``backout'' changeset as the parent of the working directory.
jerojasro@343: \interaction{backout.manual.parents}
jerojasro@343: Now we have two isolated sets of changes.
jerojasro@343: \interaction{backout.manual.heads}
jerojasro@343: 
jerojasro@343: Let's think about what we expect to see as the contents of
jerojasro@343: \filename{myfile} now.  The first change should be present, because
jerojasro@343: we've never backed it out.  The second change should be missing, as
jerojasro@343: that's the change we backed out.  Since the history graph shows the
jerojasro@343: third change as a separate head, we \emph{don't} expect to see the
jerojasro@343: third change present in \filename{myfile}.
jerojasro@343: \interaction{backout.manual.cat}
jerojasro@343: To get the third change back into the file, we just do a normal merge
jerojasro@343: of our two heads.
jerojasro@343: \interaction{backout.manual.merge}
jerojasro@343: Afterwards, the graphical history of our repository looks like
jerojasro@343: figure~\ref{fig:undo:backout-manual-merge}.
jerojasro@343: 
jerojasro@343: \begin{figure}[htb]
jerojasro@343:   \centering
jerojasro@343:   \grafix{undo-manual-merge}
jerojasro@343:   \caption{Manually merging a backout change}
jerojasro@343:   \label{fig:undo:backout-manual-merge}
jerojasro@343: \end{figure}
jerojasro@343: 
jerojasro@343: \subsection{Why \hgcmd{backout} works as it does}
jerojasro@343: 
jerojasro@343: Here's a brief description of how the \hgcmd{backout} command works.
jerojasro@343: \begin{enumerate}
jerojasro@343: \item It ensures that the working directory is ``clean'', i.e.~that
jerojasro@343:   the output of \hgcmd{status} would be empty.
jerojasro@343: \item It remembers the current parent of the working directory.  Let's
jerojasro@343:   call this changeset \texttt{orig}
jerojasro@343: \item It does the equivalent of a \hgcmd{update} to sync the working
jerojasro@343:   directory to the changeset you want to back out.  Let's call this
jerojasro@343:   changeset \texttt{backout}
jerojasro@343: \item It finds the parent of that changeset.  Let's call that
jerojasro@343:   changeset \texttt{parent}.
jerojasro@343: \item For each file that the \texttt{backout} changeset affected, it
jerojasro@343:   does the equivalent of a \hgcmdargs{revert}{-r parent} on that file,
jerojasro@343:   to restore it to the contents it had before that changeset was
jerojasro@343:   committed.
jerojasro@343: \item It commits the result as a new changeset.  This changeset has
jerojasro@343:   \texttt{backout} as its parent.
jerojasro@343: \item If you specify \hgopt{backout}{--merge} on the command line, it
jerojasro@343:   merges with \texttt{orig}, and commits the result of the merge.
jerojasro@343: \end{enumerate}
jerojasro@343: 
jerojasro@343: An alternative way to implement the \hgcmd{backout} command would be
jerojasro@343: to \hgcmd{export} the to-be-backed-out changeset as a diff, then use
jerojasro@343: the \cmdopt{patch}{--reverse} option to the \command{patch} command to
jerojasro@343: reverse the effect of the change without fiddling with the working
jerojasro@343: directory.  This sounds much simpler, but it would not work nearly as
jerojasro@343: well.
jerojasro@343: 
jerojasro@343: The reason that \hgcmd{backout} does an update, a commit, a merge, and
jerojasro@343: another commit is to give the merge machinery the best chance to do a
jerojasro@343: good job when dealing with all the changes \emph{between} the change
jerojasro@343: you're backing out and the current tip.  
jerojasro@343: 
jerojasro@343: If you're backing out a changeset that's~100 revisions back in your
jerojasro@343: project's history, the chances that the \command{patch} command will
jerojasro@343: be able to apply a reverse diff cleanly are not good, because
jerojasro@343: intervening changes are likely to have ``broken the context'' that
jerojasro@343: \command{patch} uses to determine whether it can apply a patch (if
jerojasro@343: this sounds like gibberish, see \ref{sec:mq:patch} for a
jerojasro@343: discussion of the \command{patch} command).  Also, Mercurial's merge
jerojasro@343: machinery will handle files and directories being renamed, permission
jerojasro@343: changes, and modifications to binary files, none of which
jerojasro@343: \command{patch} can deal with.
jerojasro@343: 
jerojasro@343: \section{Changes that should never have been}
jerojasro@343: \label{sec:undo:aaaiiieee}
jerojasro@343: 
jerojasro@343: Most of the time, the \hgcmd{backout} command is exactly what you need
jerojasro@343: if you want to undo the effects of a change.  It leaves a permanent
jerojasro@343: record of exactly what you did, both when committing the original
jerojasro@343: changeset and when you cleaned up after it.
jerojasro@343: 
jerojasro@343: On rare occasions, though, you may find that you've committed a change
jerojasro@343: that really should not be present in the repository at all.  For
jerojasro@343: example, it would be very unusual, and usually considered a mistake,
jerojasro@343: to commit a software project's object files as well as its source
jerojasro@343: files.  Object files have almost no intrinsic value, and they're
jerojasro@343: \emph{big}, so they increase the size of the repository and the amount
jerojasro@343: of time it takes to clone or pull changes.
jerojasro@343: 
jerojasro@343: Before I discuss the options that you have if you commit a ``brown
jerojasro@343: paper bag'' change (the kind that's so bad that you want to pull a
jerojasro@343: brown paper bag over your head), let me first discuss some approaches
jerojasro@343: that probably won't work.
jerojasro@343: 
jerojasro@343: Since Mercurial treats history as accumulative---every change builds
jerojasro@343: on top of all changes that preceded it---you generally can't just make
jerojasro@343: disastrous changes disappear.  The one exception is when you've just
jerojasro@343: committed a change, and it hasn't been pushed or pulled into another
jerojasro@343: repository.  That's when you can safely use the \hgcmd{rollback}
jerojasro@343: command, as I detailed in section~\ref{sec:undo:rollback}.
jerojasro@343: 
jerojasro@343: After you've pushed a bad change to another repository, you
jerojasro@343: \emph{could} still use \hgcmd{rollback} to make your local copy of the
jerojasro@343: change disappear, but it won't have the consequences you want.  The
jerojasro@343: change will still be present in the remote repository, so it will
jerojasro@343: reappear in your local repository the next time you pull.
jerojasro@343: 
jerojasro@343: If a situation like this arises, and you know which repositories your
jerojasro@343: bad change has propagated into, you can \emph{try} to get rid of the
jerojasro@343: changeefrom \emph{every} one of those repositories.  This is, of
jerojasro@343: course, not a satisfactory solution: if you miss even a single
jerojasro@343: repository while you're expunging, the change is still ``in the
jerojasro@343: wild'', and could propagate further.
jerojasro@343: 
jerojasro@343: If you've committed one or more changes \emph{after} the change that
jerojasro@343: you'd like to see disappear, your options are further reduced.
jerojasro@343: Mercurial doesn't provide a way to ``punch a hole'' in history,
jerojasro@343: leaving changesets intact.
jerojasro@343: 
jerojasro@343: XXX This needs filling out.  The \texttt{hg-replay} script in the
jerojasro@343: \texttt{examples} directory works, but doesn't handle merge
jerojasro@343: changesets.  Kind of an important omission.
jerojasro@343: 
jerojasro@343: \subsection{Protect yourself from ``escaped'' changes}
jerojasro@343: 
jerojasro@343: If you've committed some changes to your local repository and they've
jerojasro@343: been pushed or pulled somewhere else, this isn't necessarily a
jerojasro@343: disaster.  You can protect yourself ahead of time against some classes
jerojasro@343: of bad changeset.  This is particularly easy if your team usually
jerojasro@343: pulls changes from a central repository.
jerojasro@343: 
jerojasro@343: By configuring some hooks on that repository to validate incoming
jerojasro@343: changesets (see chapter~\ref{chap:hook}), you can automatically
jerojasro@343: prevent some kinds of bad changeset from being pushed to the central
jerojasro@343: repository at all.  With such a configuration in place, some kinds of
jerojasro@343: bad changeset will naturally tend to ``die out'' because they can't
jerojasro@343: propagate into the central repository.  Better yet, this happens
jerojasro@343: without any need for explicit intervention.
jerojasro@343: 
jerojasro@343: For instance, an incoming change hook that verifies that a changeset
jerojasro@343: will actually compile can prevent people from inadvertantly ``breaking
jerojasro@343: the build''.
jerojasro@343: 
jerojasro@343: \section{Finding the source of a bug}
jerojasro@343: \label{sec:undo:bisect}
jerojasro@343: 
jerojasro@343: While it's all very well to be able to back out a changeset that
jerojasro@343: introduced a bug, this requires that you know which changeset to back
jerojasro@343: out.  Mercurial provides an invaluable command, called
jerojasro@343: \hgcmd{bisect}, that helps you to automate this process and accomplish
jerojasro@343: it very efficiently.
jerojasro@343: 
jerojasro@343: The idea behind the \hgcmd{bisect} command is that a changeset has
jerojasro@343: introduced some change of behaviour that you can identify with a
jerojasro@343: simple binary test.  You don't know which piece of code introduced the
jerojasro@343: change, but you know how to test for the presence of the bug.  The
jerojasro@343: \hgcmd{bisect} command uses your test to direct its search for the
jerojasro@343: changeset that introduced the code that caused the bug.
jerojasro@343: 
jerojasro@343: Here are a few scenarios to help you understand how you might apply
jerojasro@343: this command.
jerojasro@343: \begin{itemize}
jerojasro@343: \item The most recent version of your software has a bug that you
jerojasro@343:   remember wasn't present a few weeks ago, but you don't know when it
jerojasro@343:   was introduced.  Here, your binary test checks for the presence of
jerojasro@343:   that bug.
jerojasro@343: \item You fixed a bug in a rush, and now it's time to close the entry
jerojasro@343:   in your team's bug database.  The bug database requires a changeset
jerojasro@343:   ID when you close an entry, but you don't remember which changeset
jerojasro@343:   you fixed the bug in.  Once again, your binary test checks for the
jerojasro@343:   presence of the bug.
jerojasro@343: \item Your software works correctly, but runs~15\% slower than the
jerojasro@343:   last time you measured it.  You want to know which changeset
jerojasro@343:   introduced the performance regression.  In this case, your binary
jerojasro@343:   test measures the performance of your software, to see whether it's
jerojasro@343:   ``fast'' or ``slow''.
jerojasro@343: \item The sizes of the components of your project that you ship
jerojasro@343:   exploded recently, and you suspect that something changed in the way
jerojasro@343:   you build your project.
jerojasro@343: \end{itemize}
jerojasro@343: 
jerojasro@343: From these examples, it should be clear that the \hgcmd{bisect}
jerojasro@343: command is not useful only for finding the sources of bugs.  You can
jerojasro@343: use it to find any ``emergent property'' of a repository (anything
jerojasro@343: that you can't find from a simple text search of the files in the
jerojasro@343: tree) for which you can write a binary test.
jerojasro@343: 
jerojasro@343: We'll introduce a little bit of terminology here, just to make it
jerojasro@343: clear which parts of the search process are your responsibility, and
jerojasro@343: which are Mercurial's.  A \emph{test} is something that \emph{you} run
jerojasro@343: when \hgcmd{bisect} chooses a changeset.  A \emph{probe} is what
jerojasro@343: \hgcmd{bisect} runs to tell whether a revision is good.  Finally,
jerojasro@343: we'll use the word ``bisect'', as both a noun and a verb, to stand in
jerojasro@343: for the phrase ``search using the \hgcmd{bisect} command.
jerojasro@343: 
jerojasro@343: One simple way to automate the searching process would be simply to
jerojasro@343: probe every changeset.  However, this scales poorly.  If it took ten
jerojasro@343: minutes to test a single changeset, and you had 10,000 changesets in
jerojasro@343: your repository, the exhaustive approach would take on average~35
jerojasro@343: \emph{days} to find the changeset that introduced a bug.  Even if you
jerojasro@343: knew that the bug was introduced by one of the last 500 changesets,
jerojasro@343: and limited your search to those, you'd still be looking at over 40
jerojasro@343: hours to find the changeset that introduced your bug.
jerojasro@343: 
jerojasro@343: What the \hgcmd{bisect} command does is use its knowledge of the
jerojasro@343: ``shape'' of your project's revision history to perform a search in
jerojasro@343: time proportional to the \emph{logarithm} of the number of changesets
jerojasro@343: to check (the kind of search it performs is called a dichotomic
jerojasro@343: search).  With this approach, searching through 10,000 changesets will
jerojasro@343: take less than three hours, even at ten minutes per test (the search
jerojasro@343: will require about 14 tests).  Limit your search to the last hundred
jerojasro@343: changesets, and it will take only about an hour (roughly seven tests).
jerojasro@343: 
jerojasro@343: The \hgcmd{bisect} command is aware of the ``branchy'' nature of a
jerojasro@343: Mercurial project's revision history, so it has no problems dealing
jerojasro@343: with branches, merges, or multiple heads in a repoository.  It can
jerojasro@343: prune entire branches of history with a single probe, which is how it
jerojasro@343: operates so efficiently.
jerojasro@343: 
jerojasro@343: \subsection{Using the \hgcmd{bisect} command}
jerojasro@343: 
jerojasro@343: Here's an example of \hgcmd{bisect} in action.
jerojasro@343: 
jerojasro@343: \begin{note}
jerojasro@343:   In versions 0.9.5 and earlier of Mercurial, \hgcmd{bisect} was not a
jerojasro@343:   core command: it was distributed with Mercurial as an extension.
jerojasro@343:   This section describes the built-in command, not the old extension.
jerojasro@343: \end{note}
jerojasro@343: 
jerojasro@343: Now let's create a repository, so that we can try out the
jerojasro@343: \hgcmd{bisect} command in isolation.
jerojasro@343: \interaction{bisect.init}
jerojasro@343: We'll simulate a project that has a bug in it in a simple-minded way:
jerojasro@343: create trivial changes in a loop, and nominate one specific change
jerojasro@343: that will have the ``bug''.  This loop creates 35 changesets, each
jerojasro@343: adding a single file to the repository.  We'll represent our ``bug''
jerojasro@343: with a file that contains the text ``i have a gub''.
jerojasro@343: \interaction{bisect.commits}
jerojasro@343: 
jerojasro@343: The next thing that we'd like to do is figure out how to use the
jerojasro@343: \hgcmd{bisect} command.  We can use Mercurial's normal built-in help
jerojasro@343: mechanism for this.
jerojasro@343: \interaction{bisect.help}
jerojasro@343: 
jerojasro@343: The \hgcmd{bisect} command works in steps.  Each step proceeds as follows.
jerojasro@343: \begin{enumerate}
jerojasro@343: \item You run your binary test.
jerojasro@343:   \begin{itemize}
jerojasro@343:   \item If the test succeeded, you tell \hgcmd{bisect} by running the
jerojasro@343:     \hgcmdargs{bisect}{good} command.
jerojasro@343:   \item If it failed, run the \hgcmdargs{bisect}{--bad} command.
jerojasro@343:   \end{itemize}
jerojasro@343: \item The command uses your information to decide which changeset to
jerojasro@343:   test next.
jerojasro@343: \item It updates the working directory to that changeset, and the
jerojasro@343:   process begins again.
jerojasro@343: \end{enumerate}
jerojasro@343: The process ends when \hgcmd{bisect} identifies a unique changeset
jerojasro@343: that marks the point where your test transitioned from ``succeeding''
jerojasro@343: to ``failing''.
jerojasro@343: 
jerojasro@343: To start the search, we must run the \hgcmdargs{bisect}{--reset} command.
jerojasro@343: \interaction{bisect.search.init}
jerojasro@343: 
jerojasro@343: In our case, the binary test we use is simple: we check to see if any
jerojasro@343: file in the repository contains the string ``i have a gub''.  If it
jerojasro@343: does, this changeset contains the change that ``caused the bug''.  By
jerojasro@343: convention, a changeset that has the property we're searching for is
jerojasro@343: ``bad'', while one that doesn't is ``good''.
jerojasro@343: 
jerojasro@343: Most of the time, the revision to which the working directory is
jerojasro@343: synced (usually the tip) already exhibits the problem introduced by
jerojasro@343: the buggy change, so we'll mark it as ``bad''.
jerojasro@343: \interaction{bisect.search.bad-init}
jerojasro@343: 
jerojasro@343: Our next task is to nominate a changeset that we know \emph{doesn't}
jerojasro@343: have the bug; the \hgcmd{bisect} command will ``bracket'' its search
jerojasro@343: between the first pair of good and bad changesets.  In our case, we
jerojasro@343: know that revision~10 didn't have the bug.  (I'll have more words
jerojasro@343: about choosing the first ``good'' changeset later.)
jerojasro@343: \interaction{bisect.search.good-init}
jerojasro@343: 
jerojasro@343: Notice that this command printed some output.
jerojasro@343: \begin{itemize}
jerojasro@343: \item It told us how many changesets it must consider before it can
jerojasro@343:   identify the one that introduced the bug, and how many tests that
jerojasro@343:   will require.
jerojasro@343: \item It updated the working directory to the next changeset to test,
jerojasro@343:   and told us which changeset it's testing.
jerojasro@343: \end{itemize}
jerojasro@343: 
jerojasro@343: We now run our test in the working directory.  We use the
jerojasro@343: \command{grep} command to see if our ``bad'' file is present in the
jerojasro@343: working directory.  If it is, this revision is bad; if not, this
jerojasro@343: revision is good.
jerojasro@343: \interaction{bisect.search.step1}
jerojasro@343: 
jerojasro@343: This test looks like a perfect candidate for automation, so let's turn
jerojasro@343: it into a shell function.
jerojasro@343: \interaction{bisect.search.mytest}
jerojasro@343: We can now run an entire test step with a single command,
jerojasro@343: \texttt{mytest}.
jerojasro@343: \interaction{bisect.search.step2}
jerojasro@343: A few more invocations of our canned test step command, and we're
jerojasro@343: done.
jerojasro@343: \interaction{bisect.search.rest}
jerojasro@343: 
jerojasro@343: Even though we had~40 changesets to search through, the \hgcmd{bisect}
jerojasro@343: command let us find the changeset that introduced our ``bug'' with
jerojasro@343: only five tests.  Because the number of tests that the \hgcmd{bisect}
jerojasro@343: command performs grows logarithmically with the number of changesets to
jerojasro@343: search, the advantage that it has over the ``brute force'' search
jerojasro@343: approach increases with every changeset you add.
jerojasro@343: 
jerojasro@343: \subsection{Cleaning up after your search}
jerojasro@343: 
jerojasro@343: When you're finished using the \hgcmd{bisect} command in a
jerojasro@343: repository, you can use the \hgcmdargs{bisect}{reset} command to drop
jerojasro@343: the information it was using to drive your search.  The command
jerojasro@343: doesn't use much space, so it doesn't matter if you forget to run this
jerojasro@343: command.  However, \hgcmd{bisect} won't let you start a new search in
jerojasro@343: that repository until you do a \hgcmdargs{bisect}{reset}.
jerojasro@343: \interaction{bisect.search.reset}
jerojasro@343: 
jerojasro@343: \section{Tips for finding bugs effectively}
jerojasro@343: 
jerojasro@343: \subsection{Give consistent input}
jerojasro@343: 
jerojasro@343: The \hgcmd{bisect} command requires that you correctly report the
jerojasro@343: result of every test you perform.  If you tell it that a test failed
jerojasro@343: when it really succeeded, it \emph{might} be able to detect the
jerojasro@343: inconsistency.  If it can identify an inconsistency in your reports,
jerojasro@343: it will tell you that a particular changeset is both good and bad.
jerojasro@343: However, it can't do this perfectly; it's about as likely to report
jerojasro@343: the wrong changeset as the source of the bug.
jerojasro@343: 
jerojasro@343: \subsection{Automate as much as possible}
jerojasro@343: 
jerojasro@343: When I started using the \hgcmd{bisect} command, I tried a few times
jerojasro@343: to run my tests by hand, on the command line.  This is an approach
jerojasro@343: that I, at least, am not suited to.  After a few tries, I found that I
jerojasro@343: was making enough mistakes that I was having to restart my searches
jerojasro@343: several times before finally getting correct results.
jerojasro@343: 
jerojasro@343: My initial problems with driving the \hgcmd{bisect} command by hand
jerojasro@343: occurred even with simple searches on small repositories; if the
jerojasro@343: problem you're looking for is more subtle, or the number of tests that
jerojasro@343: \hgcmd{bisect} must perform increases, the likelihood of operator
jerojasro@343: error ruining the search is much higher.  Once I started automating my
jerojasro@343: tests, I had much better results.
jerojasro@343: 
jerojasro@343: The key to automated testing is twofold:
jerojasro@343: \begin{itemize}
jerojasro@343: \item always test for the same symptom, and
jerojasro@343: \item always feed consistent input to the \hgcmd{bisect} command.
jerojasro@343: \end{itemize}
jerojasro@343: In my tutorial example above, the \command{grep} command tests for the
jerojasro@343: symptom, and the \texttt{if} statement takes the result of this check
jerojasro@343: and ensures that we always feed the same input to the \hgcmd{bisect}
jerojasro@343: command.  The \texttt{mytest} function marries these together in a
jerojasro@343: reproducible way, so that every test is uniform and consistent.
jerojasro@343: 
jerojasro@343: \subsection{Check your results}
jerojasro@343: 
jerojasro@343: Because the output of a \hgcmd{bisect} search is only as good as the
jerojasro@343: input you give it, don't take the changeset it reports as the
jerojasro@343: absolute truth.  A simple way to cross-check its report is to manually
jerojasro@343: run your test at each of the following changesets:
jerojasro@343: \begin{itemize}
jerojasro@343: \item The changeset that it reports as the first bad revision.  Your
jerojasro@343:   test should still report this as bad.
jerojasro@343: \item The parent of that changeset (either parent, if it's a merge).
jerojasro@343:   Your test should report this changeset as good.
jerojasro@343: \item A child of that changeset.  Your test should report this
jerojasro@343:   changeset as bad.
jerojasro@343: \end{itemize}
jerojasro@343: 
jerojasro@343: \subsection{Beware interference between bugs}
jerojasro@343: 
jerojasro@343: It's possible that your search for one bug could be disrupted by the
jerojasro@343: presence of another.  For example, let's say your software crashes at
jerojasro@343: revision 100, and worked correctly at revision 50.  Unknown to you,
jerojasro@343: someone else introduced a different crashing bug at revision 60, and
jerojasro@343: fixed it at revision 80.  This could distort your results in one of
jerojasro@343: several ways.
jerojasro@343: 
jerojasro@343: It is possible that this other bug completely ``masks'' yours, which
jerojasro@343: is to say that it occurs before your bug has a chance to manifest
jerojasro@343: itself.  If you can't avoid that other bug (for example, it prevents
jerojasro@343: your project from building), and so can't tell whether your bug is
jerojasro@343: present in a particular changeset, the \hgcmd{bisect} command cannot
jerojasro@343: help you directly.  Instead, you can mark a changeset as untested by
jerojasro@343: running \hgcmdargs{bisect}{--skip}.
jerojasro@343: 
jerojasro@343: A different problem could arise if your test for a bug's presence is
jerojasro@343: not specific enough.  If you check for ``my program crashes'', then
jerojasro@343: both your crashing bug and an unrelated crashing bug that masks it
jerojasro@343: will look like the same thing, and mislead \hgcmd{bisect}.
jerojasro@343: 
jerojasro@343: Another useful situation in which to use \hgcmdargs{bisect}{--skip} is
jerojasro@343: if you can't test a revision because your project was in a broken and
jerojasro@343: hence untestable state at that revision, perhaps because someone
jerojasro@343: checked in a change that prevented the project from building.
jerojasro@343: 
jerojasro@343: \subsection{Bracket your search lazily}
jerojasro@343: 
jerojasro@343: Choosing the first ``good'' and ``bad'' changesets that will mark the
jerojasro@343: end points of your search is often easy, but it bears a little
jerojasro@343: discussion nevertheless.  From the perspective of \hgcmd{bisect}, the
jerojasro@343: ``newest'' changeset is conventionally ``bad'', and the older
jerojasro@343: changeset is ``good''.
jerojasro@343: 
jerojasro@343: If you're having trouble remembering when a suitable ``good'' change
jerojasro@343: was, so that you can tell \hgcmd{bisect}, you could do worse than
jerojasro@343: testing changesets at random.  Just remember to eliminate contenders
jerojasro@343: that can't possibly exhibit the bug (perhaps because the feature with
jerojasro@343: the bug isn't present yet) and those where another problem masks the
jerojasro@343: bug (as I discussed above).
jerojasro@343: 
jerojasro@343: Even if you end up ``early'' by thousands of changesets or months of
jerojasro@343: history, you will only add a handful of tests to the total number that
jerojasro@343: \hgcmd{bisect} must perform, thanks to its logarithmic behaviour.
jerojasro@343: 
jerojasro@343: %%% Local Variables: 
jerojasro@343: %%% mode: latex
jerojasro@343: %%% TeX-master: "00book"
jerojasro@343: %%% End: