cg-admin-rewritehist

   1 #!/usr/bin/env bash
   2 #
   3 # Rewrite revision history
   4 # Copyright (c) Petr Baudis, 2006
   5 #
   6 # Lets you rewrite GIT revision history by creating a new branch from
   7 # your current branch by applying custom filters on each revision.
   8 # Those filters can modify each tree (e.g. removing a file or running
   9 # a perl rewrite on all files) or information about each commit.
  10 # Otherwise, all information (including original commit times or merge
  11 # information) will be preserved.
  12 #
  13 # The command takes the new branch name as a mandatory argument and
  14 # the filters as optional arguments. If you specify no filters, the
  15 # commits will be recommitted without any changes, which would normally
  16 # have no effect and result with the new branch pointing to the same
  17 # branch as your current branch. (Nevertheless, this may be useful in
  18 # the future for compensating for some Git bugs or such, therefore
  19 # such a usage is permitted.)
  20 #
  21 # WARNING! The rewritten history will have different ids for all the
  22 # objects and will not converge with the original branch. You will not
  23 # be able to easily push and distribute the rewritten branch. Please do
  24 # not use this command if you do not know the full implications, and
  25 # avoid using it anyway - do not do what a simple single commit on top
  26 # of the current version would fix.
  27 #
  28 # Always verify that the rewritten version is correct before disposing
  29 # the original branch.
  30 #
  31 # Note that since this operation is extensively I/O expensive, it might
  32 # be a good idea to do it off-disk, e.g. on tmpfs. Reportedly the speedup
  33 # is very noticeable.
  34 #
  35 # OPTIONS
  36 # -------
  37 # -d TEMPDIR:: The path to the temporary tree used for rewriting
  38 #       When applying a tree filter, the command needs to temporary
  39 #       checkout the tree to some directory, which may consume
  40 #       considerable space in case of large projects. By default it
  41 #       does this in the '.git-rewrite/' directory but you can override
  42 #       that choice by this parameter.
  43 #
  44 # -r STARTREV:: The commit id to start the rewrite at
  45 #       Normally, the command will rewrite the entire history. If you
  46 #       pass this argument, though, this will be the first commit it
  47 #       will rewrite and keep the previous commits intact.
  48 #
  49 # -k KEEPREV:: A commit id until which _not_ to rewrite history
  50 #       If you pass this argument, this commit and all of its
  51 #       predecessors are kept intact.
  52 #
  53 # Filters
  54 # ~~~~~~~
  55 # The filters are applied in the order as listed below. The COMMAND
  56 # argument is always evaluated in shell using the 'eval' command.
  57 # The $GIT_COMMIT environment variable is permanently set to contain
  58 # the id of the commit being rewritten. The author/committer environment
  59 # variables are set before the first filter is run.
  60 #
  61 # A 'map' function is available that takes an "original sha1 id" argument
  62 # and outputs a "rewritten sha1 id" if the commit has been already
  63 # rewritten, fails otherwise; the 'map' function can return several
  64 # ids on separate lines if your commit filter emitted multiple commits
  65 # (see below).
  66 #
  67 # --env-filter COMMAND:: The filter for modifying environment
  68 #       This is the filter for modifying the environment in which
  69 #       the commit will be performed. Specifically, you might want
  70 #       to rewrite the author/committer name/email/time environment
  71 #       variables (see `cg-commit` for details). Do not forget to
  72 #       re-export the variables.
  73 #
  74 # --tree-filter COMMAND:: The filter for rewriting tree (and its contents)
  75 #       This is the filter for rewriting the tree and its contents.
  76 #       The COMMAND argument is evaluated in shell with the working
  77 #       directory set to the root of the checked out tree. The new tree
  78 #       is then used as-is (new files are auto-added, disappeared files
  79 #       are auto-removed - .gitignore files nor any other ignore rules
  80 #       HAVE NO EFFECT!).
  81 #
  82 # --index-filter COMMAND:: The filter for rewriting index
  83 #       This is the filter for rewriting the Git's directory index.
  84 #       It is similar to the tree filter but does not check out the
  85 #       tree, which makes it much faster. However, you must use the
  86 #       lowlevel Git index manipulation commands to do your work.
  87 #
  88 # --parent-filter COMMAND:: The filter for rewriting parents
  89 #       This is the filter for rewriting the commit's parent list.
  90 #       It will receive the parent string on stdin and shall output
  91 #       the new parent string on stdout. The parent string is in
  92 #       format accepted by `git-commit-tree`: empty for initial
  93 #       commit, "-p parent" for a normal commit and "-p parent1
  94 #       -p parent2 -p parent3 ..." for a merge commit.
  95 #
  96 # --msg-filter COMMAND:: The filter for rewriting commit message
  97 #       This is the filter for rewriting the commit messages.
  98 #       The COMMAND argument is evaluated in shell with the original
  99 #       commit message on standard input; its standard output is
 100 #       is used as the new commit message.
 101 #
 102 # --commit-filter COMMAND:: The filter for performing the commit
 103 #       If this filter is passed, it will be called instead of the
 104 #       `git-commit-tree` command, with those arguments:
 105 #
 106 #               TREE_ID [-p PARENT_COMMIT_ID]...
 107 #
 108 #       and the log message on stdin. The commit id is expected on
 109 #       stdout. As a special extension, the commit filter may emit
 110 #       multiple commit ids; in that case, all of them will be used
 111 #       as parents instead of the original commit in further commits.
 112 #
 113 # --tag-name-filter COMMAND:: The filter for rewriting tag names.
 114 #       If this filter is passed, it will be called for every tag ref
 115 #       that points to a rewritten object (or to a tag object which
 116 #       points to a rewritten object). The original tag name is passed
 117 #       via standard input, and the new tag name is expected on standard
 118 #       output.
 119 #
 120 #       The original tags are not deleted, but can be overwritten;
 121 #       use "--tag-name-filter=cat" to simply update the tags. In this
 122 #       case, be very careful and make sure you have the old tags
 123 #       backed up in case the conversion has run afoul.
 124 #
 125 #       Note that there is currently no support for proper rewriting of
 126 #       tag objects; in layman terms, if the tag has a message or signature
 127 #       attached, the rewritten tag won't have it. Sorry. (It is by
 128 #       definition impossible to preserve signatures at any rate, though.)
 129 #
 130 # EXAMPLE USAGE
 131 # -------------
 132 # Suppose you want to remove a file (containing confidential information
 133 # or copyright violation) from all commits:
 134 #
 135 #       cg-admin-rewritehist --tree-filter 'rm filename' newbranch
 136 #
 137 # A significantly faster version:
 138 #
 139 #       cg-admin-rewritehist --index-filter 'git-update-index --remove filename' newbranch
 140 #
 141 # Now, you will get the rewritten history saved in the branch 'newbranch'
 142 # (your current branch is left untouched).
 143 #
 144 # To "etch-graft" a commit to the revision history (set a commit to be
 145 # the parent of the current initial commit and propagate that):
 146 #
 147 #       cg-admin-rewritehist --parent-filter sed\ 's/^$/-p graftcommitid/' newbranch
 148 #
 149 # (if the parent string is empty - therefore we are dealing with the
 150 # initial commit - add graftcommit as a parent). Note that this assumes
 151 # history with a single root (that is, no cg-merge -j happened). If this
 152 # is not the case, use:
 153 #
 154 #       cg-admin-rewritehist --parent-filter 'cat; [ "$GIT_COMMIT" = "COMMIT" ] && echo "-p GRAFTCOMMIT"' newbranch
 155 #
 156 # To remove commits authored by "Darl McBribe" from the history:
 157 #
 158 #       cg-admin-rewritehist --commit-filter 'if [ "$GIT_AUTHOR_NAME" = "Darl McBribe" ]; then shift; while [ -n "$1" ]; do shift; echo "$1"; shift; done; else git-commit-tree "$@"; fi' newbranch
 159 #
 160 # (the shift magic first throws away the tree id and then the -p
 161 # parameters). Note that this handles merges properly! In case Darl
 162 # committed a merge between P1 and P2, it will be propagated properly
 163 # and all children of the merge will become merge commits with P1,P2
 164 # as their parents instead of the merge commit.
 165 #
 166 # To restrict rewriting to only part of the history, use -r or -k or both.
 167 # Consider this history:
 168 #
 169 #            D--E--F--G--H
 170 #           /     /
 171 #       A--B-----C
 172 #
 173 # To rewrite only commits F,G,H, use:
 174 #
 175 #       cg-admin-rewritehist -r F ...
 176 #
 177 # To rewrite commits E,F,G,H, use one of these:
 178 #
 179 #       cg-admin-rewritehist -r E -k C ...
 180 #       cg-admin-rewritehist -k D -k C ...
 181
 182 # Testsuite: TODO
 183
 184 set -e
 185
 186 USAGE="cg-admin-rewritehist [-d TEMPDIR] [-r STARTREV]... [-k KEEPREV]... [FILTERS] DESTBRANCH"
 187 _git_wc_unneeded=1
 188 _git_requires_root=1
 189
 190 . "${COGITO_LIB}"cg-Xlib || exit 1
 191
 192
 193 map()
 194 {
 195         [ -r "$workdir/../map/$1" ] || return 1
 196         cat "$workdir/../map/$1"
 197 }
 198
 199
 200 tempdir=.git-rewrite
 201 declare -a unchanged
 202 filter_env=
 203 filter_tree=
 204 filter_index=
 205 filter_parent=
 206 filter_msg=cat
 207 filter_commit='git-commit-tree "$@"'
 208 filter_tag_name=
 209 while optparse; do
 210         if optparse -d=; then
 211                 tempdir="$OPTARG"
 212         elif optparse -r=; then
 213                 unchanged=(${unchanged[@]} $(cg-object-id -p "$OPTARG")) || exit 1
 214         elif optparse -k=; then
 215                 unchanged=(${unchanged[@]} $(cg-object-id -c "$OPTARG")) || exit 1
 216         elif optparse --env-filter=; then
 217                 filter_env="$OPTARG"
 218         elif optparse --tree-filter=; then
 219                 filter_tree="$OPTARG"
 220         elif optparse --index-filter=; then
 221                 filter_index="$OPTARG"
 222         elif optparse --parent-filter=; then
 223                 filter_parent="$OPTARG"
 224         elif optparse --msg-filter=; then
 225                 filter_msg="$OPTARG"
 226         elif optparse --commit-filter=; then
 227                 filter_commit="$OPTARG"
 228         elif optparse --tag-name-filter=; then
 229                 filter_tag_name="$OPTARG"
 230         else
 231                 optfail
 232         fi
 233 done
 234
 235 dstbranch="${ARGS[0]}"
 236 [ -n "$dstbranch" ] || die "missing branch name"
 237 ! exists_ref "refs/heads/$dstbranch" || die "branch $dstbranch already exists"
 238 [ ! -s "$_git/branches/$dstbranch" ] || die "branch $dstbranch is already a remote branch"
 239
 240 [ ! -e "$tempdir" ] || die "$tempdir already exists, please remove it"
 241 mkdir -p "$tempdir/t"
 242 cd "$tempdir/t"
 243 workdir="$(pwd)"
 244
 245 [ -n "$GIT_DIR" ] || export GIT_DIR=.git
 246 [[ "$GIT_DIR" == /* ]] || export GIT_DIR="$(pwd)/../../$GIT_DIR"
 247 export GIT_INDEX_FILE="$(pwd)/../index"
 248 git-read-tree # seed the index file
 249
 250 ret=0
 251
 252
 253 mkdir ../map # map old->new commit ids for rewriting parents
 254
 255 # seed with identity mappings for the parents where we start off
 256 for commit in ${unchanged[@]}; do
 257         echo $commit > ../map/$commit
 258 done
 259
 260 git-rev-list --topo-order HEAD ${unchanged[@]/#/^} | tac >../revs
 261 commits=$(cat ../revs | wc -l)
 262
 263 if [ $commits -eq 0 ]; then
 264         die "Found nothing to rewrite"
 265 fi
 266
 267 i=0
 268 while read commit; do
 269         i=$((i+1))
 270         echo -n "$commit ($i/$commits) "
 271
 272         git-read-tree -i -m $commit
 273
 274         export GIT_COMMIT=$commit
 275         git-cat-file commit "$commit" >../commit
 276
 277         eval "$(pick_author <../commit)"
 278         eval "$(pick_id committer COMMITTER <../commit)"
 279         eval "$filter_env"
 280
 281         if [ "$filter_tree" ]; then
 282                 git-checkout-index -f -u -a
 283                 # files that $commit removed are now still in the working tree;
 284                 # remove them, else they would be added again
 285                 git-ls-files -z --others | xargs -0 rm -f
 286                 eval "$filter_tree"
 287                 git-diff-index -r $commit | cut -f 2- | tr '\n' '\0' | \
 288                         xargs -0 git-update-index --add --replace --remove
 289                 git-ls-files -z --others | \
 290                         xargs -0 git-update-index --add --replace --remove
 291         fi
 292
 293         eval "$filter_index"
 294
 295         parentstr=
 296         for parent in $(cg-object-id -p $commit); do
 297                 if [ -r "../map/$parent" ]; then
 298                         for reparent in $(cat "../map/$parent"); do
 299                                 parentstr="$parentstr -p $reparent"
 300                         done
 301                 else
 302                         die "assertion failed: parent $parent for commit $commit not found in rewritten ones"
 303                 fi
 304         done
 305         if [ "$filter_parent" ]; then
 306                 parentstr="$(echo "$parentstr" | eval "$filter_parent")"
 307         fi
 308
 309         sed -e '1,/^$/d' <../commit | \
 310                 eval "$filter_msg" | \
 311                 sh -c "$filter_commit" git-commit-tree $(git-write-tree) $parentstr | \
 312                 tee ../map/$commit
 313 done <../revs
 314
 315 git-update-ref refs/heads/"$dstbranch" $(head -n 1 ../map/$(tail -n 1 ../revs))
 316 if [ "$(cat ../map/$(tail -n 1 ../revs) | wc -l)" -gt 1 ]; then
 317         echo "WARNING: Your commit filter caused the head commit to expand to several rewritten commits. Only the first such commit was recorded as the current $dstbranch head but you will need to resolve the situation now (probably by manually merging the other commits). These are all the commits:" >&2
 318         sed 's/^/       /' ../map/$(tail -n 1 ../revs) >&2
 319         ret=1
 320 fi
 321
 322 if [ "$filter_tag_name" ]; then
 323         git-for-each-ref --format='%(objectname) %(objecttype) %(refname)' refs/tags |
 324         while read sha1 type ref; do
 325                 ref="${ref#refs/tags/}"
 326                 # XXX: Rewrite tagged trees as well?
 327                 if [ "$type" != "commit" -a "$type" != "tag" ]; then
 328                         continue;
 329                 fi
 330
 331                 if [ "$type" = "tag" ]; then
 332                         # Dereference to a commit
 333                         sha1t="$sha1"
 334                         sha1="$(cg-object-id -c "$sha1" 2>/dev/null)" || continue
 335                 fi
 336
 337                 [ -f "../map/$sha1" ] || continue
 338                 new_sha1="$(cat "../map/$sha1")"
 339                 export GIT_COMMIT="$sha1"
 340                 new_ref="$(echo "$ref" | eval "$filter_tag_name")"
 341
 342                 echo "$ref -> $new_ref ($sha1 -> $new_sha1)"
 343
 344                 if [ "$type" = "tag" ]; then
 345                         # Warn that we are not rewriting the tag object itself.
 346                         warn "unreferencing tag object $sha1t"
 347                 fi
 348
 349                 git-update-ref "refs/tags/$new_ref" "$new_sha1"
 350         done
 351 fi
 352
 353 cd ../..
 354 rm -rf "$tempdir"
 355 echo "Rewritten history saved to the $dstbranch branch"
 356
 357 exit $ret