From: Paul Mackerras <paulus@samba.org>
Date: Mon, 6 Feb 2006 22:13:52 +0000 (+1100)
Subject: gitk: Use git-diff-tree --cc for showing the diffs for merges
X-Git-Tag: v1.2.0~40^2
X-Git-Url: https://git.verplant.org/?a=commitdiff_plain;h=b77b02785d2f589ca336fc449ee1c27837263dac;p=git.git

gitk: Use git-diff-tree --cc for showing the diffs for merges

This replaces a lot of code that used the result from several 2-way
diffs to generate a combined diff for a merge.  Now we just use
git-diff-tree --cc and colorize the output a bit, which is a lot
simpler, and has the enormous advantage that if the diff doesn't
show quite what someone thinks it should show, I can deflect the
blame to someone else. :)

Signed-off-by: Paul Mackerras <paulus@samba.org>
---

diff --git a/gitk b/gitk
index 9ba3d16b..e4821406 100755
--- a/gitk
+++ b/gitk
@@ -542,8 +542,19 @@ proc makewindow {rargs} {
     $ctext tag conf m2 -fore green
     $ctext tag conf m3 -fore purple
     $ctext tag conf m4 -fore brown
+    $ctext tag conf m5 -fore "#009090"
+    $ctext tag conf m6 -fore magenta
+    $ctext tag conf m7 -fore "#808000"
+    $ctext tag conf m8 -fore "#009000"
+    $ctext tag conf m9 -fore "#ff0080"
+    $ctext tag conf m10 -fore cyan
+    $ctext tag conf m11 -fore "#b07070"
+    $ctext tag conf m12 -fore "#70b0f0"
+    $ctext tag conf m13 -fore "#70f0b0"
+    $ctext tag conf m14 -fore "#f0b070"
+    $ctext tag conf m15 -fore "#ff70b0"
     $ctext tag conf mmax -fore darkgrey
-    set mergemax 5
+    set mergemax 16
     $ctext tag conf mresult -font [concat $textfont bold]
     $ctext tag conf msep -font [concat $textfont bold]
     $ctext tag conf found -back yellow
@@ -2182,6 +2193,7 @@ proc selectline {l isnew} {
     global canvy0 linespc parents nparents children
     global cflist currentid sha1entry
     global commentend idtags idline linknum
+    global mergemax
 
     $canv delete hover
     normalline
@@ -2265,11 +2277,26 @@ proc selectline {l isnew} {
     }
  
     set comment {}
-    if {[info exists parents($id)]} {
+    if {$nparents($id) > 1} {
+	set np 0
 	foreach p $parents($id) {
-	    append comment "Parent: [commit_descriptor $p]\n"
+	    if {$np >= $mergemax} {
+		set tag mmax
+	    } else {
+		set tag m$np
+	    }
+	    $ctext insert end "Parent: " $tag
+	    appendwithlinks [commit_descriptor $p]
+	    incr np
+	}
+    } else {
+	if {[info exists parents($id)]} {
+	    foreach p $parents($id) {
+		append comment "Parent: [commit_descriptor $p]\n"
+	    }
 	}
     }
+
     if {[info exists children($id)]} {
 	foreach c $children($id) {
 	    append comment "Child:  [commit_descriptor $c]\n"
@@ -2361,529 +2388,100 @@ proc goforw {} {
 }
 
 proc mergediff {id} {
-    global parents diffmergeid diffmergegca mergefilelist diffpindex
+    global parents diffmergeid diffopts mdifffd
+    global difffilestart
 
     set diffmergeid $id
-    set diffpindex -1
-    set diffmergegca [findgca $parents($id)]
-    if {[info exists mergefilelist($id)]} {
-	if {$mergefilelist($id) ne {}} {
-	    showmergediff
-	}
-    } else {
-	contmergediff {}
-    }
-}
-
-proc findgca {ids} {
-    set gca {}
-    foreach id $ids {
-	if {$gca eq {}} {
-	    set gca $id
-	} else {
-	    if {[catch {
-		set gca [exec git-merge-base $gca $id]
-	    } err]} {
-		return {}
-	    }
-	}
-    }
-    return $gca
-}
-
-proc contmergediff {ids} {
-    global diffmergeid diffpindex parents nparents diffmergegca
-    global treediffs mergefilelist diffids treepending
-
-    # diff the child against each of the parents, and diff
-    # each of the parents against the GCA.
-    while 1 {
-	if {[lindex $ids 1] == $diffmergeid && $diffmergegca ne {}} {
-	    set ids [list $diffmergegca [lindex $ids 0]]
-	} else {
-	    if {[incr diffpindex] >= $nparents($diffmergeid)} break
-	    set p [lindex $parents($diffmergeid) $diffpindex]
-	    set ids [list $p $diffmergeid]
-	}
-	if {![info exists treediffs($ids)]} {
-	    set diffids $ids
-	    if {![info exists treepending]} {
-		gettreediffs $ids
-	    }
-	    return
-	}
-    }
-
-    # If a file in some parent is different from the child and also
-    # different from the GCA, then it's interesting.
-    # If we don't have a GCA, then a file is interesting if it is
-    # different from the child in all the parents.
-    if {$diffmergegca ne {}} {
-	set files {}
-	foreach p $parents($diffmergeid) {
-	    set gcadiffs $treediffs([list $diffmergegca $p])
-	    foreach f $treediffs([list $p $diffmergeid]) {
-		if {[lsearch -exact $files $f] < 0
-		    && [lsearch -exact $gcadiffs $f] >= 0} {
-		    lappend files $f
-		}
-	    }
-	}
-	set files [lsort $files]
-    } else {
-	set p [lindex $parents($diffmergeid) 0]
-	set files $treediffs([list $diffmergeid $p])
-	for {set i 1} {$i < $nparents($diffmergeid) && $files ne {}} {incr i} {
-	    set p [lindex $parents($diffmergeid) $i]
-	    set df $treediffs([list $p $diffmergeid])
-	    set nf {}
-	    foreach f $files {
-		if {[lsearch -exact $df $f] >= 0} {
-		    lappend nf $f
-		}
-	    }
-	    set files $nf
-	}
-    }
-
-    set mergefilelist($diffmergeid) $files
-    if {$files ne {}} {
-	showmergediff
-    }
-}
-
-proc showmergediff {} {
-    global cflist diffmergeid mergefilelist parents
-    global diffopts diffinhunk currentfile currenthunk filelines
-    global diffblocked groupfilelast mergefds groupfilenum grouphunks
-
-    set files $mergefilelist($diffmergeid)
-    foreach f $files {
-	$cflist insert end $f
-    }
+    catch {unset difffilestart}
+    # this doesn't seem to actually affect anything...
     set env(GIT_DIFF_OPTS) $diffopts
-    set flist {}
-    catch {unset currentfile}
-    catch {unset currenthunk}
-    catch {unset filelines}
-    catch {unset groupfilenum}
-    catch {unset grouphunks}
-    set groupfilelast -1
-    foreach p $parents($diffmergeid) {
-	set cmd [list | git-diff-tree -p $p $diffmergeid]
-	set cmd [concat $cmd $mergefilelist($diffmergeid)]
-	if {[catch {set f [open $cmd r]} err]} {
-	    error_popup "Error getting diffs: $err"
-	    foreach f $flist {
-		catch {close $f}
-	    }
-	    return
-	}
-	lappend flist $f
-	set ids [list $diffmergeid $p]
-	set mergefds($ids) $f
-	set diffinhunk($ids) 0
-	set diffblocked($ids) 0
-	fconfigure $f -blocking 0
-	fileevent $f readable [list getmergediffline $f $ids $diffmergeid]
+    set cmd [concat | git-diff-tree --no-commit-id --cc $id]
+    if {[catch {set mdf [open $cmd r]} err]} {
+	error_popup "Error getting merge diffs: $err"
+	return
     }
+    fconfigure $mdf -blocking 0
+    set mdifffd($id) $mdf
+    fileevent $mdf readable [list getmergediffline $mdf $id]
+    set nextupdate [expr {[clock clicks -milliseconds] + 100}]
 }
 
-proc getmergediffline {f ids id} {
-    global diffmergeid diffinhunk diffoldlines diffnewlines
-    global currentfile currenthunk
-    global diffoldstart diffnewstart diffoldlno diffnewlno
-    global diffblocked mergefilelist
-    global noldlines nnewlines difflcounts filelines
+proc getmergediffline {mdf id} {
+    global diffmergeid ctext cflist nextupdate nparents mergemax
+    global difffilestart
 
-    set n [gets $f line]
+    set n [gets $mdf line]
     if {$n < 0} {
-	if {![eof $f]} return
-    }
-
-    if {!([info exists diffmergeid] && $diffmergeid == $id)} {
-	if {$n < 0} {
-	    close $f
+	if {[eof $mdf]} {
+	    close $mdf
 	}
 	return
     }
-
-    if {$diffinhunk($ids) != 0} {
-	set fi $currentfile($ids)
-	if {$n > 0 && [regexp {^[-+ \\]} $line match]} {
-	    # continuing an existing hunk
-	    set line [string range $line 1 end]
-	    set p [lindex $ids 1]
-	    if {$match eq "-" || $match eq " "} {
-		set filelines($p,$fi,$diffoldlno($ids)) $line
-		incr diffoldlno($ids)
-	    }
-	    if {$match eq "+" || $match eq " "} {
-		set filelines($id,$fi,$diffnewlno($ids)) $line
-		incr diffnewlno($ids)
-	    }
-	    if {$match eq " "} {
-		if {$diffinhunk($ids) == 2} {
-		    lappend difflcounts($ids) \
-			[list $noldlines($ids) $nnewlines($ids)]
-		    set noldlines($ids) 0
-		    set diffinhunk($ids) 1
-		}
-		incr noldlines($ids)
-	    } elseif {$match eq "-" || $match eq "+"} {
-		if {$diffinhunk($ids) == 1} {
-		    lappend difflcounts($ids) [list $noldlines($ids)]
-		    set noldlines($ids) 0
-		    set nnewlines($ids) 0
-		    set diffinhunk($ids) 2
-		}
-		if {$match eq "-"} {
-		    incr noldlines($ids)
-		} else {
-		    incr nnewlines($ids)
-		}
-	    }
-	    # and if it's \ No newline at end of line, then what?
-	    return
-	}
-	# end of a hunk
-	if {$diffinhunk($ids) == 1 && $noldlines($ids) != 0} {
-	    lappend difflcounts($ids) [list $noldlines($ids)]
-	} elseif {$diffinhunk($ids) == 2
-		  && ($noldlines($ids) != 0 || $nnewlines($ids) != 0)} {
-	    lappend difflcounts($ids) [list $noldlines($ids) $nnewlines($ids)]
-	}
-	set currenthunk($ids) [list $currentfile($ids) \
-				   $diffoldstart($ids) $diffnewstart($ids) \
-				   $diffoldlno($ids) $diffnewlno($ids) \
-				   $difflcounts($ids)]
-	set diffinhunk($ids) 0
-	# -1 = need to block, 0 = unblocked, 1 = is blocked
-	set diffblocked($ids) -1
-	processhunks
-	if {$diffblocked($ids) == -1} {
-	    fileevent $f readable {}
-	    set diffblocked($ids) 1
-	}
-    }
-
-    if {$n < 0} {
-	# eof
-	if {!$diffblocked($ids)} {
-	    close $f
-	    set currentfile($ids) [llength $mergefilelist($diffmergeid)]
-	    set currenthunk($ids) [list $currentfile($ids) 0 0 0 0 {}]
-	    processhunks
-	}
-    } elseif {[regexp {^diff --git a/(.*) b/} $line match fname]} {
-	# start of a new file
-	set currentfile($ids) \
-	    [lsearch -exact $mergefilelist($diffmergeid) $fname]
-    } elseif {[regexp {^@@ -([0-9]+),([0-9]+) \+([0-9]+),([0-9]+) @@(.*)} \
-		   $line match f1l f1c f2l f2c rest]} {
-	if {[info exists currentfile($ids)] && $currentfile($ids) >= 0} {
-	    # start of a new hunk
-	    if {$f1l == 0 && $f1c == 0} {
-		set f1l 1
-	    }
-	    if {$f2l == 0 && $f2c == 0} {
-		set f2l 1
-	    }
-	    set diffinhunk($ids) 1
-	    set diffoldstart($ids) $f1l
-	    set diffnewstart($ids) $f2l
-	    set diffoldlno($ids) $f1l
-	    set diffnewlno($ids) $f2l
-	    set difflcounts($ids) {}
-	    set noldlines($ids) 0
-	    set nnewlines($ids) 0
-	}
-    }
-}
-
-proc processhunks {} {
-    global diffmergeid parents nparents currenthunk
-    global mergefilelist diffblocked mergefds
-    global grouphunks grouplinestart grouplineend groupfilenum
-
-    set nfiles [llength $mergefilelist($diffmergeid)]
-    while 1 {
-	set fi $nfiles
-	set lno 0
-	# look for the earliest hunk
-	foreach p $parents($diffmergeid) {
-	    set ids [list $diffmergeid $p]
-	    if {![info exists currenthunk($ids)]} return
-	    set i [lindex $currenthunk($ids) 0]
-	    set l [lindex $currenthunk($ids) 2]
-	    if {$i < $fi || ($i == $fi && $l < $lno)} {
-		set fi $i
-		set lno $l
-		set pi $p
-	    }
-	}
-
-	if {$fi < $nfiles} {
-	    set ids [list $diffmergeid $pi]
-	    set hunk $currenthunk($ids)
-	    unset currenthunk($ids)
-	    if {$diffblocked($ids) > 0} {
-		fileevent $mergefds($ids) readable \
-		    [list getmergediffline $mergefds($ids) $ids $diffmergeid]
-	    }
-	    set diffblocked($ids) 0
-
-	    if {[info exists groupfilenum] && $groupfilenum == $fi
-		&& $lno <= $grouplineend} {
-		# add this hunk to the pending group
-		lappend grouphunks($pi) $hunk
-		set endln [lindex $hunk 4]
-		if {$endln > $grouplineend} {
-		    set grouplineend $endln
-		}
-		continue
-	    }
-	}
-
-	# succeeding stuff doesn't belong in this group, so
-	# process the group now
-	if {[info exists groupfilenum]} {
-	    processgroup
-	    unset groupfilenum
-	    unset grouphunks
-	}
-
-	if {$fi >= $nfiles} break
-
-	# start a new group
-	set groupfilenum $fi
-	set grouphunks($pi) [list $hunk]
-	set grouplinestart $lno
-	set grouplineend [lindex $hunk 4]
+    if {![info exists diffmergeid] || $id != $diffmergeid} {
+	return
     }
-}
-
-proc processgroup {} {
-    global groupfilelast groupfilenum difffilestart
-    global mergefilelist diffmergeid ctext filelines
-    global parents diffmergeid diffoffset
-    global grouphunks grouplinestart grouplineend nparents
-    global mergemax
-
     $ctext conf -state normal
-    set id $diffmergeid
-    set f $groupfilenum
-    if {$groupfilelast != $f} {
+    if {[regexp {^diff --cc (.*)} $line match fname]} {
+	# start of a new file
 	$ctext insert end "\n"
 	set here [$ctext index "end - 1c"]
-	set difffilestart($f) $here
-	set mark fmark.[expr {$f + 1}]
-	$ctext mark set $mark $here
-	$ctext mark gravity $mark left
-	set header [lindex $mergefilelist($id) $f]
-	set l [expr {(78 - [string length $header]) / 2}]
+	set i [$cflist index end]
+	$ctext mark set fmark.$i $here
+	$ctext mark gravity fmark.$i left
+	set difffilestart([expr {$i-1}]) $here
+	$cflist insert end $fname
+	set l [expr {(78 - [string length $fname]) / 2}]
 	set pad [string range "----------------------------------------" 1 $l]
-	$ctext insert end "$pad $header $pad\n" filesep
-	set groupfilelast $f
-	foreach p $parents($id) {
-	    set diffoffset($p) 0
-	}
-    }
-
-    $ctext insert end "@@" msep
-    set nlines [expr {$grouplineend - $grouplinestart}]
-    set events {}
-    set pnum 0
-    foreach p $parents($id) {
-	set startline [expr {$grouplinestart + $diffoffset($p)}]
-	set ol $startline
-	set nl $grouplinestart
-	if {[info exists grouphunks($p)]} {
-	    foreach h $grouphunks($p) {
-		set l [lindex $h 2]
-		if {$nl < $l} {
-		    for {} {$nl < $l} {incr nl} {
-			set filelines($p,$f,$ol) $filelines($id,$f,$nl)
-			incr ol
-		    }
-		}
-		foreach chunk [lindex $h 5] {
-		    if {[llength $chunk] == 2} {
-			set olc [lindex $chunk 0]
-			set nlc [lindex $chunk 1]
-			set nnl [expr {$nl + $nlc}]
-			lappend events [list $nl $nnl $pnum $olc $nlc]
-			incr ol $olc
-			set nl $nnl
-		    } else {
-			incr ol [lindex $chunk 0]
-			incr nl [lindex $chunk 0]
-		    }
-		}
-	    }
-	}
-	if {$nl < $grouplineend} {
-	    for {} {$nl < $grouplineend} {incr nl} {
-		set filelines($p,$f,$ol) $filelines($id,$f,$nl)
-		incr ol
-	    }
-	}
-	set nlines [expr {$ol - $startline}]
-	$ctext insert end " -$startline,$nlines" msep
-	incr pnum
-    }
-
-    set nlines [expr {$grouplineend - $grouplinestart}]
-    $ctext insert end " +$grouplinestart,$nlines @@\n" msep
-
-    set events [lsort -integer -index 0 $events]
-    set nevents [llength $events]
-    set nmerge $nparents($diffmergeid)
-    set l $grouplinestart
-    for {set i 0} {$i < $nevents} {set i $j} {
-	set nl [lindex $events $i 0]
-	while {$l < $nl} {
-	    $ctext insert end " $filelines($id,$f,$l)\n"
-	    incr l
-	}
-	set e [lindex $events $i]
-	set enl [lindex $e 1]
-	set j $i
-	set active {}
-	while 1 {
-	    set pnum [lindex $e 2]
-	    set olc [lindex $e 3]
-	    set nlc [lindex $e 4]
-	    if {![info exists delta($pnum)]} {
-		set delta($pnum) [expr {$olc - $nlc}]
-		lappend active $pnum
+	$ctext insert end "$pad $fname $pad\n" filesep
+    } elseif {[regexp {^@@} $line]} {
+	$ctext insert end "$line\n" hunksep
+    } elseif {[regexp {^[0-9a-f]{40}$} $line] || [regexp {^index} $line]} {
+	# do nothing
+    } else {
+	# parse the prefix - one ' ', '-' or '+' for each parent
+	set np $nparents($id)
+	set spaces {}
+	set minuses {}
+	set pluses {}
+	set isbad 0
+	for {set j 0} {$j < $np} {incr j} {
+	    set c [string range $line $j $j]
+	    if {$c == " "} {
+		lappend spaces $j
+	    } elseif {$c == "-"} {
+		lappend minuses $j
+	    } elseif {$c == "+"} {
+		lappend pluses $j
 	    } else {
-		incr delta($pnum) [expr {$olc - $nlc}]
-	    }
-	    if {[incr j] >= $nevents} break
-	    set e [lindex $events $j]
-	    if {[lindex $e 0] >= $enl} break
-	    if {[lindex $e 1] > $enl} {
-		set enl [lindex $e 1]
-	    }
-	}
-	set nlc [expr {$enl - $l}]
-	set ncol mresult
-	set bestpn -1
-	if {[llength $active] == $nmerge - 1} {
-	    # no diff for one of the parents, i.e. it's identical
-	    for {set pnum 0} {$pnum < $nmerge} {incr pnum} {
-		if {![info exists delta($pnum)]} {
-		    if {$pnum < $mergemax} {
-			lappend ncol m$pnum
-		    } else {
-			lappend ncol mmax
-		    }
-		    break
-		}
-	    }
-	} elseif {[llength $active] == $nmerge} {
-	    # all parents are different, see if one is very similar
-	    set bestsim 30
-	    for {set pnum 0} {$pnum < $nmerge} {incr pnum} {
-		set sim [similarity $pnum $l $nlc $f \
-			     [lrange $events $i [expr {$j-1}]]]
-		if {$sim > $bestsim} {
-		    set bestsim $sim
-		    set bestpn $pnum
-		}
-	    }
-	    if {$bestpn >= 0} {
-		lappend ncol m$bestpn
+		set isbad 1
+		break
 	    }
 	}
-	set pnum -1
-	foreach p $parents($id) {
-	    incr pnum
-	    if {![info exists delta($pnum)] || $pnum == $bestpn} continue
-	    set olc [expr {$nlc + $delta($pnum)}]
-	    set ol [expr {$l + $diffoffset($p)}]
-	    incr diffoffset($p) $delta($pnum)
-	    unset delta($pnum)
-	    for {} {$olc > 0} {incr olc -1} {
-		$ctext insert end "-$filelines($p,$f,$ol)\n" m$pnum
-		incr ol
-	    }
+	set tags {}
+	set num {}
+	if {!$isbad && $minuses ne {} && $pluses eq {}} {
+	    # line doesn't appear in result, parents in $minuses have the line
+	    set num [lindex $minuses 0]
+	} elseif {!$isbad && $pluses ne {} && $minuses eq {}} {
+	    # line appears in result, parents in $pluses don't have the line
+	    lappend tags mresult
+	    set num [lindex $spaces 0]
 	}
-	set endl [expr {$l + $nlc}]
-	if {$bestpn >= 0} {
-	    # show this pretty much as a normal diff
-	    set p [lindex $parents($id) $bestpn]
-	    set ol [expr {$l + $diffoffset($p)}]
-	    incr diffoffset($p) $delta($bestpn)
-	    unset delta($bestpn)
-	    for {set k $i} {$k < $j} {incr k} {
-		set e [lindex $events $k]
-		if {[lindex $e 2] != $bestpn} continue
-		set nl [lindex $e 0]
-		set ol [expr {$ol + $nl - $l}]
-		for {} {$l < $nl} {incr l} {
-		    $ctext insert end "+$filelines($id,$f,$l)\n" $ncol
-		}
-		set c [lindex $e 3]
-		for {} {$c > 0} {incr c -1} {
-		    $ctext insert end "-$filelines($p,$f,$ol)\n" m$bestpn
-		    incr ol
-		}
-		set nl [lindex $e 1]
-		for {} {$l < $nl} {incr l} {
-		    $ctext insert end "+$filelines($id,$f,$l)\n" mresult
-		}
+	if {$num ne {}} {
+	    if {$num >= $mergemax} {
+		set num "max"
 	    }
+	    lappend tags m$num
 	}
-	for {} {$l < $endl} {incr l} {
-	    $ctext insert end "+$filelines($id,$f,$l)\n" $ncol
-	}
-    }
-    while {$l < $grouplineend} {
-	$ctext insert end " $filelines($id,$f,$l)\n"
-	incr l
+	$ctext insert end "$line\n" $tags
     }
     $ctext conf -state disabled
-}
-
-proc similarity {pnum l nlc f events} {
-    global diffmergeid parents diffoffset filelines
-
-    set id $diffmergeid
-    set p [lindex $parents($id) $pnum]
-    set ol [expr {$l + $diffoffset($p)}]
-    set endl [expr {$l + $nlc}]
-    set same 0
-    set diff 0
-    foreach e $events {
-	if {[lindex $e 2] != $pnum} continue
-	set nl [lindex $e 0]
-	set ol [expr {$ol + $nl - $l}]
-	for {} {$l < $nl} {incr l} {
-	    incr same [string length $filelines($id,$f,$l)]
-	    incr same
-	}
-	set oc [lindex $e 3]
-	for {} {$oc > 0} {incr oc -1} {
-	    incr diff [string length $filelines($p,$f,$ol)]
-	    incr diff
-	    incr ol
-	}
-	set nl [lindex $e 1]
-	for {} {$l < $nl} {incr l} {
-	    incr diff [string length $filelines($id,$f,$l)]
-	    incr diff
-	}
-    }
-    for {} {$l < $endl} {incr l} {
-	incr same [string length $filelines($id,$f,$l)]
-	incr same
-    }
-    if {$same == 0} {
-	return 0
+    if {[clock clicks -milliseconds] >= $nextupdate} {
+	incr nextupdate 100
+	fileevent $mdf readable {}
+	update
+	fileevent $mdf readable [list getmergediffline $mdf $id]
     }
-    return [expr {200 * $same / (2 * $same + $diff)}]
 }
 
 proc startdiff {ids} {