Hi,
I'm trying to understand what is the difference between different merge
strategies. git-merge-stupid is something I failed to understand from
the git history/code/documentation, so I'm asking here.
1) From git history:
It seems git-merge-stupid was created by 2276aa6 when Junio renamed
-resolve to -stupid and let -resolve use read-tree. Actually
git show 2276aa6:git-merge-stupid.sh
says -stupid uses read-tree as well.
2) From code:
It seems -stupid is better than -resolve when there are multiple bases.
-resolve just passes all bases to read-tree, while -stupid tries to find
the best one. It does it by counting unmerged entries and the less one
is a better.
Here is what I tried:
A - B - C
X \
\ D - E - F
(Where 'A' adds 'a.c' while the contents 'a', etc.)
$ git reset --hard e
$ git merge -s resolve c
Trying simple merge.
Merge made by resolve.
So it seems resolve does not completely fail if there are multiple
bases, either.
I would like to write a testcase that ensures git-merge-stupid really
picks the best base, but I don't know exactly in what situation can the
number of unmerged entires differ.
3) From documentation:
Actually -stupid is missing from Documentation/merge-strategies.txt. I
plan to send a patch to add it, once I understnad what it does. :-)
Thanks.
[ Writing to Linus as suggested by Dscho, adding Junio to CC as -
according to git log - both strategy created by him. ]
I'm trying to understand what is the difference between different merge
strategies. git-merge-stupid is something I failed to understand from
the git history/code/documentation, so I'm asking here.
1) From git history:
It seems git-merge-stupid was created by 2276aa6 when Junio renamed
-resolve to -stupid and let -resolve use read-tree. Actually
git show 2276aa6:git-merge-stupid.sh
says -stupid uses read-tree as well.
I think -stupid should probably be removed.
The history of -stupid is from doing the simple single-tree resolve that
git-read-tree can do, but then doing the obvious hack of just trying to
pick the base that gives the least number of conflicts.
HOWEVER.
- in practice, there's seldom any actual point to it. In 99% of all
cases, you only have a single merge base anyway.
- if you have a workflow that encourages criss-cross merges (which makes
the above "in practice" not be true), the common case will be that the
merge base doesn't much matter.
- Counting conflicts by looking at the numbe of files that conflict is a
pretty stupid metric anyway. Yes, it's obvious, and yes, I bet there
are cases where it does the right thing, but I also bet there are cases
where it does the _wrong_ thing - it might pick a merge base with fewer
files conflicting, but with harder conflicts.
- the "recursive" merge strategy simply handles things better. There's
not really any reason to use a "pick random merge base that happens to
give least conflicts", when the recursive strategy does something much
more natural.
So you shouldn't really compare -stupid to -resolve. You should compare
-stupid to -recursive, and the latter is simply much better.
2) From code:
It seems -stupid is better than -resolve when there are multiple bases.
Maybe. And maybe not.
$ git merge -s resolve c
Trying simple merge.
Merge made by resolve.
So it seems resolve does not completely fail if there are multiple
bases, either.
I think -resolve can handle up to 6 bases, or something like that. After
that it should fail with a "I cannot read more than 8 trees" or something
(eight being the two trees to be merged, plus the six bases).
And with multiple bases, it will already pick the best one on a per-file
basis (I think - I should know the threeway merge, but it is pretty
confusing code) rather than trying to pick one globally. Not pretty, but
it's yet another reason why -stupid is actually stupid, and not worth it.
So -stupid in _theory_ can handle cases that -resolve cannot (more than
six bases), but (a) that doesn't happen and (b) you'd be better off with
-recursive anyway.
3) From documentation:
Actually -stupid is missing from Documentation/merge-strategies.txt. I
plan to send a patch to add it, once I understnad what it does. :-)
Well, see above. I think there's a reason why -stupid isn't even worth
documenting. It might be better off just removed.
Linus
As pointed out by Linus, this strategy tries to take the best merge
base, but 'recursive' just does it better. If one needs something more
than 'resolve' then he/she should really use 'recursive' and not
'stupid'.
---
On Thu, Jul 03, 2008 at 10:08:54AM -0700, Linus Torvalds [off-list ref] wrote:
I think -stupid should probably be removed.
Here is an attempt to do so.
The history of -stupid is from doing the simple single-tree resolve
that
git-read-tree can do, but then doing the obvious hack of just trying
to
pick the base that gives the least number of conflicts.
HOWEVER.
(...)
@@ -1,80 +0,0 @@-#!/bin/sh-#-# Copyright (c) 2005 Linus Torvalds-#-# Resolve two trees, 'stupid merge'.--# The first parameters up to -- are merge bases; the rest are heads.-bases= head= remotes= sep_seen=-for arg-do- case ",$sep_seen,$head,$arg," in- *,--,)- sep_seen=yes- ;;- ,yes,,*)- head=$arg- ;;- ,yes,*)- remotes="$remotes$arg "- ;;- *)- bases="$bases$arg "- ;;- esac-done--# Give up if we are given two or more remotes -- not handling octopus.-case "$remotes" in-?*' '?*)- exit 2 ;;-esac--# Find an optimum merge base if there are more than one candidates.-case "$bases" in-?*' '?*)- echo "Trying to find the optimum merge base."- G=.tmp-index$$- best=- best_cnt=-1- for c in $bases- do- rm -f $G- GIT_INDEX_FILE=$G git read-tree -m $c $head $remotes \- 2>/dev/null || continue- # Count the paths that are unmerged.- cnt=`GIT_INDEX_FILE=$G git ls-files --unmerged | wc -l`- if test $best_cnt -le 0 -o $cnt -le $best_cnt- then- best=$c- best_cnt=$cnt- if test "$best_cnt" -eq 0- then- # Cannot do any better than all trivial merge.- break- fi- fi- done- rm -f $G- common="$best"- ;;-*)- common="$bases"- ;;-esac--git update-index --refresh 2>/dev/null-git read-tree -u -m $common $head $remotes || exit 2-echo "Trying simple merge."-if result_tree=$(git write-tree 2>/dev/null)-then- exit 0-else- echo "Simple merge failed, trying Automatic merge."- if git-merge-index -o git-merge-one-file -a- then- exit 0- else- exit 1- fi-fi
As pointed out by Linus, this strategy tries to take the best merge
base, but 'recursive' just does it better. If one needs something more
than 'resolve' then he/she should really use 'recursive' and not
'stupid'.
Given that it may still serve as a good example, don't remove it, just
move it to contrib/examples.
---
On Fri, Jul 04, 2008 at 02:07:01AM +0200, Miklos Vajna [off-list ref] wrote:
On Thu, Jul 03, 2008 at 04:54:31PM -0700, Junio C Hamano
[off-list ref] wrote:
quoted
It is called stupid for a reason ;-).
It has been sitting there as an example for a long time, and I do
not think anybody minds removing it.
OK, then should I resend a patch that moves it to contrib/examples?
@@ -0,0 +1,80 @@+#!/bin/sh+#+# Copyright (c) 2005 Linus Torvalds+#+# Resolve two trees, 'stupid merge'.++# The first parameters up to -- are merge bases; the rest are heads.+bases=head=remotes=sep_seen=+forarg+do+case",$sep_seen,$head,$arg,"in+*,--,)+sep_seen=yes+;;+,yes,,*)+head=$arg+;;+,yes,*)+remotes="$remotes$arg "+;;+*)+bases="$bases$arg "+;;+esac+done++# Give up if we are given two or more remotes -- not handling octopus.+case"$remotes"in+?*' '?*)+exit2;;+esac++# Find an optimum merge base if there are more than one candidates.+case"$bases"in+?*' '?*)+echo"Trying to find the optimum merge base."+G=.tmp-index$$+best=+best_cnt=-1+forcin$bases+do+rm-f$G+GIT_INDEX_FILE=$Ggitread-tree-m$c$head$remotes\+2>/dev/null||continue+# Count the paths that are unmerged.+cnt=`GIT_INDEX_FILE=$Ggitls-files--unmerged|wc-l`+iftest$best_cnt-le0-o$cnt-le$best_cnt+then+best=$c+best_cnt=$cnt+iftest"$best_cnt"-eq0+then+# Cannot do any better than all trivial merge.+break+fi+fi+done+rm-f$G+common="$best"+;;+*)+common="$bases"+;;+esac++gitupdate-index--refresh2>/dev/null+gitread-tree-u-m$common$head$remotes||exit2+echo"Trying simple merge."+ifresult_tree=$(gitwrite-tree2>/dev/null)+then+exit0+else+echo"Simple merge failed, trying Automatic merge."+ifgit-merge-index-ogit-merge-one-file-a+then+exit0+else+exit1+fi+fi
@@ -1,80 +0,0 @@-#!/bin/sh-#-# Copyright (c) 2005 Linus Torvalds-#-# Resolve two trees, 'stupid merge'.--# The first parameters up to -- are merge bases; the rest are heads.-bases= head= remotes= sep_seen=-for arg-do- case ",$sep_seen,$head,$arg," in- *,--,)- sep_seen=yes- ;;- ,yes,,*)- head=$arg- ;;- ,yes,*)- remotes="$remotes$arg "- ;;- *)- bases="$bases$arg "- ;;- esac-done--# Give up if we are given two or more remotes -- not handling octopus.-case "$remotes" in-?*' '?*)- exit 2 ;;-esac--# Find an optimum merge base if there are more than one candidates.-case "$bases" in-?*' '?*)- echo "Trying to find the optimum merge base."- G=.tmp-index$$- best=- best_cnt=-1- for c in $bases- do- rm -f $G- GIT_INDEX_FILE=$G git read-tree -m $c $head $remotes \- 2>/dev/null || continue- # Count the paths that are unmerged.- cnt=`GIT_INDEX_FILE=$G git ls-files --unmerged | wc -l`- if test $best_cnt -le 0 -o $cnt -le $best_cnt- then- best=$c- best_cnt=$cnt- if test "$best_cnt" -eq 0- then- # Cannot do any better than all trivial merge.- break- fi- fi- done- rm -f $G- common="$best"- ;;-*)- common="$bases"- ;;-esac--git update-index --refresh 2>/dev/null-git read-tree -u -m $common $head $remotes || exit 2-echo "Trying simple merge."-if result_tree=$(git write-tree 2>/dev/null)-then- exit 0-else- echo "Simple merge failed, trying Automatic merge."- if git-merge-index -o git-merge-one-file -a- then- exit 0- else- exit 1- fi-fi
As pointed out by Linus, this strategy tries to take the best merge
base, but 'recursive' just does it better. If one needs something more
than 'resolve' then he/she should really use 'recursive' and not
'stupid'.
Given that it may still serve as a good example, don't remove it, just
move it to contrib/examples.
---
n Sat, Jul 05, 2008 at 04:43:51PM +0200, Miklos Vajna [off-list ref] wrote:
diff --git a/git-merge-stupid.sh b/contrib/examples/git-merge-stupid.shsimilarity index 100%rename from git-merge-stupid.shrename to contrib/examples/git-merge-stupid.sh
--
1.5.6.1.322.ge904b.dirty