Documentation / technical / rerere.txton commit Merge branch 'cb/windows-manifest' (9b9b24b)
   1Rerere
   2======
   3
   4This document describes the rerere logic.
   5
   6Conflict normalization
   7----------------------
   8
   9To ensure recorded conflict resolutions can be looked up in the rerere
  10database, even when branches are merged in a different order,
  11different branches are merged that result in the same conflict, or
  12when different conflict style settings are used, rerere normalizes the
  13conflicts before writing them to the rerere database.
  14
  15Different conflict styles and branch names are normalized by stripping
  16the labels from the conflict markers, and removing the common ancestor
  17version from the `diff3` conflict style. Branches that are merged
  18in different order are normalized by sorting the conflict hunks.  More
  19on each of those steps in the following sections.
  20
  21Once these two normalization operations are applied, a conflict ID is
  22calculated based on the normalized conflict, which is later used by
  23rerere to look up the conflict in the rerere database.
  24
  25Removing the common ancestor version
  26~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  27
  28Say we have three branches AB, AC and AC2.  The common ancestor of
  29these branches has a file with a line containing the string "A" (for
  30brevity this is called "line A" in the rest of the document).  In
  31branch AB this line is changed to "B", in AC, this line is changed to
  32"C", and branch AC2 is forked off of AC, after the line was changed to
  33"C".
  34
  35Forking a branch ABAC off of branch AB and then merging AC into it, we
  36get a conflict like the following:
  37
  38    <<<<<<< HEAD
  39    B
  40    =======
  41    C
  42    >>>>>>> AC
  43
  44Doing the analogous with AC2 (forking a branch ABAC2 off of branch AB
  45and then merging branch AC2 into it), using the diff3 conflict style,
  46we get a conflict like the following:
  47
  48    <<<<<<< HEAD
  49    B
  50    ||||||| merged common ancestors
  51    A
  52    =======
  53    C
  54    >>>>>>> AC2
  55
  56By resolving this conflict, to leave line D, the user declares:
  57
  58    After examining what branches AB and AC did, I believe that making
  59    line A into line D is the best thing to do that is compatible with
  60    what AB and AC wanted to do.
  61
  62As branch AC2 refers to the same commit as AC, the above implies that
  63this is also compatible what AB and AC2 wanted to do.
  64
  65By extension, this means that rerere should recognize that the above
  66conflicts are the same.  To do this, the labels on the conflict
  67markers are stripped, and the common ancestor version is removed.  The above
  68examples would both result in the following normalized conflict:
  69
  70    <<<<<<<
  71    B
  72    =======
  73    C
  74    >>>>>>>
  75
  76Sorting hunks
  77~~~~~~~~~~~~~
  78
  79As before, lets imagine that a common ancestor had a file with line A
  80its early part, and line X in its late part.  And then four branches
  81are forked that do these things:
  82
  83    - AB: changes A to B
  84    - AC: changes A to C
  85    - XY: changes X to Y
  86    - XZ: changes X to Z
  87
  88Now, forking a branch ABAC off of branch AB and then merging AC into
  89it, and forking a branch ACAB off of branch AC and then merging AB
  90into it, would yield the conflict in a different order.  The former
  91would say "A became B or C, what now?" while the latter would say "A
  92became C or B, what now?"
  93
  94As a reminder, the act of merging AC into ABAC and resolving the
  95conflict to leave line D means that the user declares:
  96
  97    After examining what branches AB and AC did, I believe that
  98    making line A into line D is the best thing to do that is
  99    compatible with what AB and AC wanted to do.
 100
 101So the conflict we would see when merging AB into ACAB should be
 102resolved the same way---it is the resolution that is in line with that
 103declaration.
 104
 105Imagine that similarly previously a branch XYXZ was forked from XY,
 106and XZ was merged into it, and resolved "X became Y or Z" into "X
 107became W".
 108
 109Now, if a branch ABXY was forked from AB and then merged XY, then ABXY
 110would have line B in its early part and line Y in its later part.
 111Such a merge would be quite clean.  We can construct 4 combinations
 112using these four branches ((AB, AC) x (XY, XZ)).
 113
 114Merging ABXY and ACXZ would make "an early A became B or C, a late X
 115became Y or Z" conflict, while merging ACXY and ABXZ would make "an
 116early A became C or B, a late X became Y or Z".  We can see there are
 1174 combinations of ("B or C", "C or B") x ("X or Y", "Y or X").
 118
 119By sorting, the conflict is given its canonical name, namely, "an
 120early part became B or C, a late part becames X or Y", and whenever
 121any of these four patterns appear, and we can get to the same conflict
 122and resolution that we saw earlier.
 123
 124Without the sorting, we'd have to somehow find a previous resolution
 125from combinatorial explosion.
 126
 127Conflict ID calculation
 128~~~~~~~~~~~~~~~~~~~~~~~
 129
 130Once the conflict normalization is done, the conflict ID is calculated
 131as the sha1 hash of the conflict hunks appended to each other,
 132separated by <NUL> characters.  The conflict markers are stripped out
 133before the sha1 is calculated.  So in the example above, where we
 134merge branch AC which changes line A to line C, into branch AB, which
 135changes line A to line C, the conflict ID would be
 136SHA1('B<NUL>C<NUL>').
 137
 138If there are multiple conflicts in one file, the sha1 is calculated
 139the same way with all hunks appended to each other, in the order in
 140which they appear in the file, separated by a <NUL> character.
 141
 142Nested conflicts
 143~~~~~~~~~~~~~~~~
 144
 145Nested conflicts are handled very similarly to "simple" conflicts.
 146Similar to simple conflicts, the conflict is first normalized by
 147stripping the labels from conflict markers, stripping the common ancestor
 148version, and the sorting the conflict hunks, both for the outer and the
 149inner conflict.  This is done recursively, so any number of nested
 150conflicts can be handled.
 151
 152Note that this only works for conflict markers that "cleanly nest".  If
 153there are any unmatched conflict markers, rerere will fail to handle
 154the conflict and record a conflict resolution.
 155
 156The only difference is in how the conflict ID is calculated.  For the
 157inner conflict, the conflict markers themselves are not stripped out
 158before calculating the sha1.
 159
 160Say we have the following conflict for example:
 161
 162    <<<<<<< HEAD
 163    1
 164    =======
 165    <<<<<<< HEAD
 166    3
 167    =======
 168    2
 169    >>>>>>> branch-2
 170    >>>>>>> branch-3~
 171
 172After stripping out the labels of the conflict markers, and sorting
 173the hunks, the conflict would look as follows:
 174
 175    <<<<<<<
 176    1
 177    =======
 178    <<<<<<<
 179    2
 180    =======
 181    3
 182    >>>>>>>
 183    >>>>>>>
 184
 185and finally the conflict ID would be calculated as:
 186`sha1('1<NUL><<<<<<<\n3\n=======\n2\n>>>>>>><NUL>')`