rawcommit dirstate tweak
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
rawcommit dirstate tweak
Before this patch, rawcommit can mess up the dirstate unless it is
committing against the current parent.
This patch changes rawcommit, such that when adding a child to some
node other than the current parent, rawcommit does not attempt update
the current dirstate.
This seems easily debatable; it creates an asymmetric behavior for
rawcommit. It means that when doing a rawcommit against the current
parent, there's effectively an implied "hg update" to the newly
created node. When doing a rawcommit against any other node, no such
"hg update" occurs.
The other obvious alternates would be:
1) rawcommit never update the dirstate
2) rawcommit always does an "hg update"...
This patch also includes a test for various uses of rawcommit...
Michael Fetterman
manifest hash: 428517d82a02501f14b0d8fac064411980780e91
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)
iD8DBQFCuymPywK+sNU5EO8RAvdvAKCxW1QZtyOviNfuwO592IaKApwvEACfdrYD
83m/o8oJvRKu3yGvNGHtwfk=
=KbmU
-----END PGP SIGNATURE-----
#!/bin/bash
#
# This is an example of using HGEDITOR to automate the signing of
# commits and so on.
T1=""; T2=""
cleanup_exit() {
rm -f "$T1" "$T2"
exit $1
}
case "${EDITOR:=vi}" in
emacs)
EDITOR="$EDITOR -nw"
;;
gvim|vim)
EDITOR="$EDITOR -f -o"
;;
esac
if grep -q "^HG: merge resolve" "$1" ; then
# we don't sign merges
exec $EDITOR "$1"
else
T1=`mktemp`; T2=`mktemp`
MANIFEST=`grep '^HG: manifest hash' "$1" | cut -b 19-`
echo -e "\n\nmanifest hash: $MANIFEST" >> "$T1"
grep -vE '^(HG: manifest hash .*)?$' "$1" >> "$T1"
hg diff >> "$T2"
CHECKSUM=`md5sum "$T1"`
$EDITOR "$T1" "$T2" || cleanup_exit $?
echo "$CHECKSUM" | md5sum -c 2>/dev/null && cleanup_exit 0
{
head -1 "$T1"
echo
grep -v "^HG:" "$T1" | gpg -a -u "${HGUSER:-$EMAIL}" --clearsign
} > "$T2" && mv "$T2" "$1"
cleanup_exit $?
fi