source: BLFS/libs/func_dependencies@ 28546ae

ablfs-more legacy trunk
Last change on this file since 28546ae was 28546ae, checked in by Pierre Labastie <pierre@…>, 6 years ago

Output some information during the cleaning phase in BLFS dependency engine
for the impatient

  • Property mode set to 100644
File size: 19.6 KB
Line 
1#!/bin/bash
2#
3# $Id$
4#
5
6#-----------------------------------------------------------------------------#
7# This is a set of (recursive) functions for manipulating a dependency graph. #
8# We use algorithms and definitions from chapter 4 (mainly section 4.2) of #
9# https://algs4.cs.princeton.edu/. The graph we manipulate is the directed #
10# graph of the dependencies: nodes are packages in the BLFS book. A node A is #
11# connected to a node B if package A depends on B. A topological order (rev- #
12# erted) is exactly what we want for a build order. But a topological order #
13# only exists if the graph is acyclic. We'll therefore have to remove cycles. #
14# There are a number of other features we want to consider: #
15# - edges are weighted according to the dependency requirement: #
16# 1 for required #
17# 2 for recommended #
18# 3 for optional #
19# 4 for external #
20# We should consider only edges with weight lower or equal to that #
21# specified by the user, but see below. #
22# - we do not want to build the whole book. The user specifies a set of #
23# packages, and we have to consider only nodes reachable from this set #
24# using edges of weight not exceeding the specified weight. #
25# - when doing the topological sort, we want to consider all the edges and #
26# not only those not exceeding the specified weight: If a package A in the #
27# reachable subgraph depends optionally on another package B in the same #
28# subgraph, we want to build B before A. But this means we'll have to #
29# remove cycles for all weights. #
30# - dependencies have another qualifier: before or after. The problem: if a #
31# package A depends on B with an "after" qualifier, and a package C depends #
32# on A with a "before" qualifier, C may need B to be able to use A. So the #
33# only safe way to consider "after" qualifiers is to consider that they are #
34# "before" deps for any parent of the packages considered. #
35# We'll therefore have a 3 pass procedure. First build the set of nodes #
36# reachable from the root set. Second, remove dangling edges (those pointing #
37# to packages outside the node set), and move "after" edges to "before" edges #
38# originating from the parents. Third remove cycles and generate a #
39# topological sort. #
40# #
41# TODO: document each pass
42# Data layout: #
43# TODO: needs also to document the .tree files, and the "f" qualifier
44# #
45# A node of the tree is represented by a text file <nodeName>.dep. Each edge #
46# starting from this node is represented by a line in this file. We keep #
47# those files in the same directory. We introduce a special node named root, #
48# whose edges point to the list of nodes requested by the user. Each line #
49# contains three fields: #
50# - the weight of the edge #
51# - the "before" (b) or "after" (a) qualifier #
52# - the name of the destination of the edge #
53# #
54# TODO: The following is obsolete
55# Circular dependencies: #
56# #
57# In case we find a cirdular dependency, it has the form : #
58# parent->dependency_0->...->dependency_n->dependency_0 #
59# If we want to build dependency_n before dependency_0, no problem: #
60# we just prune the tree at dependency_n. If we want to build first #
61# dependency_0, we need to put dependency_n as a dependency of parent, #
62# then erase and rebuild the subtree from there. Now, we may have met #
63# another circular dependency in the subtree, and erasing the tree makes #
64# us forget the decision which was made. So, after first generating the #
65# list of dependencies from packages.xml, we keep the generated list in #
66# a file <nodeName>.odep, which we modify according to the decision which #
67# was made. #
68#---------------------------------------------------------------------------#
69
70# Global variables:
71# A string of spaces for indenting:
72declare -a spaceSTR=" "
73# When we are backing up from a circular dependency, `parentNode'
74# contains the node which has an edge entering the cycle
75declare parentNode
76
77#---------------------#
78generate_subgraph() { #
79#---------------------#
80: <<inline_doc
81 function: Create a subgraph of all the nodes reachable from the node
82 represented by the file whose name is $1. The edges considered
83 are those with maximal weight DEP_LEVEL (recursive function).
84 input vars: $1 : file name corresponding to the node whose edges will be
85 followed for the DFS
86 $2 : weight of the edge leading to this node
87 $3 : depth (root is 1)
88 $4 : qualifier (a for after, b for before)
89 externals: vars: DEP_LEVEL contains 1 if we want to build the
90 tree only for required dependencies,
91 2 if we want also recommended ones,
92 3 if we want also optional ones, but only
93 for the requested packages,
94 4 if we want all the dependencies
95 MAIL_SERVER contains the name of the MTA we want to use.
96 files: ../xsl/dependencies.xsl: stylesheet for creating the
97 .dep files
98 ../packages.xml: File containing packages id
99 and dependencies
100 returns: 0 if the tree has been successfully created
101 output: files: for each node reachable from $1, a file <node>.dep.
102 on error: nothing
103 on success: nothing
104inline_doc
105
106local depFile=$1
107local -i weight=$2
108local -i depth=$3
109local qualifier=$4
110local -i spacing=0
111local priostring
112local buildstring
113local id_of_dep
114local prio_of_dep
115local build_of_dep
116local dep_level
117
118if (( depth < 10 )); then spacing=1; fi
119case $weight in
120 1) priostring=required ;;
121 2) priostring=recommended ;;
122 3) priostring=optional ;;
123esac
124case $qualifier in
125 a) buildstring=runtime ;;
126 b) buildstring= ;;
127esac
128dep_level=$DEP_LEVEL
129if [ "$dep_level" = 3 ] && [ "$depth" -gt 2 ]; then dep_level=2; fi
130if [ "$dep_level" -gt 3 ]; then dep_level=3; fi
131echo -en "\nNode: $depth${spaceSTR:0:$(( depth + spacing ))}${RED}${depFile%.dep}${OFF} $priostring $buildstring"
132
133depth=$(( depth + 1 ))
134if (( depth < 10 )); then spacing=1; else spacing=0; fi
135# Start of loop
136{
137while read prio_of_dep build_of_dep id_of_dep; do
138case $prio_of_dep in
139 1) priostring=required ;;
140 2) priostring=recommended ;;
141 3) priostring=optional ;;
142 4) priostring=external ;;
143esac
144case $build_of_dep in
145 a) buildstring=runtime ;;
146 b) buildstring= ;;
147esac
148# Has this entry already been seen
149 if [ -f ${id_of_dep}.dep ] ; then
150# Just display it and proceed.
151 echo -en "\nEdge: $depth${spaceSTR:0:$((depth + spacing))}${MAGENTA}${id_of_dep}${OFF} $priostring $buildstring"
152 continue
153 fi
154# Is the weight higher than requested?
155 if [ "$prio_of_dep" -gt $dep_level ]; then
156# Just display it and proceed.
157 echo -en "\n Out: $depth${spaceSTR:0:$((depth + spacing))}${YELLOW}${id_of_dep}${OFF} $priostring $buildstring"
158 continue
159 fi
160# Otherwise, let's build the corresponding subgraph.
161 xsltproc --stringparam idofdep "$id_of_dep" \
162 --stringparam MTA "$MAIL_SERVER" \
163 -o ${id_of_dep}.dep \
164 ../xsl/dependencies.xsl ../packages.xml
165
166 if [[ -s ${id_of_dep}.dep ]]; then # this dependency has dependencies
167 generate_subgraph ${id_of_dep}.dep $prio_of_dep $depth $build_of_dep
168 else # id_of_dep has no dependencies, just touch the file and display
169 touch ${id_of_dep}.dep
170 echo -en "\nLeaf: $depth${spaceSTR:0:$((depth + spacing))}${CYAN}${id_of_dep}${OFF} $priostring $buildstring"
171 fi
172done
173} <$depFile
174depth=$(( depth - 1 ))
175if (( depth < 10 )); then spacing=1; else spacing=0; fi
176echo -en "\n End: $depth${spaceSTR:0:$((depth + spacing))}${GREEN}${depFile%.dep}${OFF}"
177return 0
178}
179
180#-----------#
181path_to() { #
182#-----------#
183: <<inline_doc
184 function: check whether there is a path from $1 to $2 on the graph
185 input vars: $1 contains the filename of the starting node.
186 $2 contains the name of the node to reach
187 $3 contains the weight above which we do not want to
188 follow an edge
189 "seen" (global) contains the list of already seen nodes
190 returns: 0 if the node has been found
191 1 if not
192 on error: nothing
193 on success: nothing
194inline_doc
195local start=$1
196local seek=$2
197local prio=$3
198local prio_of_dep
199local build_of_dep
200local id_of_dep
201local r
202
203if test "${start%.dep}" = "$seek"; then return 0; fi
204seen="$seen${start%.dep} "
205if test -s $start; then
206 {
207 while read prio_of_dep build_of_dep id_of_dep; do
208 if test "$prio" -lt "$prio_of_dep"; then continue; fi
209 if ! test "${seen% $id_of_dep *}" = "$seen"; then continue; fi
210 if path_to ${id_of_dep}.dep $seek $prio; then return 0; fi
211 done
212 } < $start
213fi
214return 1
215}
216#------------------#
217clean_subgraph() { #
218#------------------#
219: <<inline_doc
220 function: Remove dangling edges and create groups of deps for "after"
221 deps: A-before->B-after->C becomes:
222 A -before-> Bgroupxx -before-> B
223 \
224 -before-> C
225 the name of the group is chosen so that it is unlikely as
226 a package name (so that it is removed when building the
227 xml book).
228 input vars: None
229 files: <node>.dep files containing dangling edges and
230 "after" qualifiers
231 returns: 0
232 output: files: <node>.dep files containing no dangling edges and
233 no "after" qualifiers
234 on error: nothing
235 on success: nothing
236inline_doc
237
238local node
239local id_of_dep
240local prio_of_dep
241local build_of_dep
242local lines_to_remove
243local lines_to_change
244local parent
245local p
246local b
247local start
248local seen
249
250for node in $(ls *.dep); do
251 if test $node = root.dep; then continue; fi
252 echo Cleaning $node
253 lines_to_remove=
254 {
255 while read prio_of_dep build_of_dep id_of_dep; do
256 if ! test -f ${id_of_dep}.dep; then
257 lines_to_remove="$lines_to_remove $id_of_dep"
258 continue
259 fi
260 done
261 } <$node
262 for id_of_dep in $lines_to_remove; do
263 sed "/\ $id_of_dep\$/d" -i $node
264 done
265done
266for node in $(grep -l ' a ' *.dep); do
267 lines_to_remove=
268 echo Process "runtime" deps in $node
269 if ! [ -e ${node%.dep}groupxx.dep ]; then
270 b=0
271 for parent in $(grep -l ${node%.dep}\$ *); do
272 p=0
273 for start in $(grep ' a ' $node | cut -d' ' -f3); do
274 seen=" "
275 if path_to ${start}.dep ${parent%.dep} 3; then p=1; break; fi
276 done
277 if test $p = 0; then
278 b=1
279 sed -i "s/\ ${node%.dep}\$/&groupxx/" $parent
280 fi
281 done
282 echo "1 b ${node%.dep}" > ${node%.dep}groupxx.dep
283 if test $b = 0; then echo "1 b ${node%.dep}groupxx" >> root.dep; fi
284 fi
285 {
286 while read prio_of_dep build_of_dep id_of_dep; do
287 if test $build_of_dep = a; then
288 if ! grep -q ${id_of_dep} ${node%.dep}groupxx.dep; then
289 echo "$prio_of_dep b ${id_of_dep}" >> ${node%.dep}groupxx.dep
290 fi
291 lines_to_remove="$lines_to_remove $id_of_dep"
292 fi
293 done
294 } <$node
295 for id_of_dep in $lines_to_remove; do
296 sed "/\ $id_of_dep\$/d" -i $node
297 done
298done
299for node in $(grep -l ' f ' *); do
300 echo Process "first" deps in $node
301 lines_to_change=
302 {
303 while read prio_of_dep build_of_dep id_of_dep; do
304 if test $build_of_dep = f; then
305 if ! test -f ${id_of_dep}-pass1.dep; then
306 cp ${id_of_dep}{,-pass1}.dep;
307 fi
308 lines_to_change="$lines_to_change $id_of_dep"
309 unset lr
310 {
311 while read p b start; do
312 seen=" "
313 if path_to ${start}.dep ${node%.dep} $p; then
314 lr="$lr $start"
315 fi
316 done
317 } < ${id_of_dep}-pass1.dep
318 for p in $lr; do
319 sed "/\ $p\$/d" -i ${id_of_dep}-pass1.dep
320 done
321 fi
322 done
323 } <$node
324 for id_of_dep in $lines_to_change; do
325 sed "/\ $id_of_dep\$/"'{s/[[:digit:]] f /1 b /;s/$/-pass1/}' -i $node
326 done
327done
328} # End clean_subgraph
329
330#----------------------------#
331generate_dependency_tree() { #
332#----------------------------#
333: <<inline_doc
334 function: Create a subtree of the dependency tree
335 (recursive function)
336 input vars: $1 : file with a list of targets (child nodes)
337 the first line of the file is the link
338 $2 : priority (1=req, 2=rec, 3=opt)
339 returns: 0 if the tree has been successfully created
340 1 if we are backing up to the parent of a circular dep
341 and there are only required deps in the cycle
342 2 if we are backing up to the parent of a circular dep
343 and there are recommended deps and no optional deps in the
344 cycle
345 3 if we are backing up to the parent of a circular dep
346 and there are optiional deps in the cycle
347 modifies: vars: ParentNode is set when return is not 0
348 output: files: for each <pkg> with dependencies in $1,
349 a file <pkg>.tree and its dependencies
350 on error: nothing
351 on success: nothing
352inline_doc
353
354local depFile=$1
355local priority=$2
356local -a rootlink
357local -a priolink
358local -a otherlink
359local -i depth
360local -i count=0
361local id_of_dep
362local build_of_dep
363local prio_of_dep
364local parent
365local lines_to_remove=
366local srootlink
367local priostring
368local dpriostring
369local i
370
371{
372read -a rootlink
373depth=${#rootlink[*]}
374read -a priolink
375srootlink="${rootlink[*]} "
376case $priority in
377 1) priostring=required ;;
378 2) priostring=recommended ;;
379 3) priostring=optional ;;
380esac
381# start of depFile
382echo -en "\nNode: $depth${spaceSTR:0:$depth}${RED}${depFile%.tree}${OFF} $priostring"
383
384while read prio_of_dep build_of_dep id_of_dep; do
385 case $prio_of_dep in
386 1) dpriostring=required ;;
387 2) dpriostring=recommended ;;
388 3) dpriostring=optional ;;
389 esac
390# count entries in file
391 (( count++ ))
392# Has this entry already been seen?
393 if [ -f ${id_of_dep}.tree ]; then # found ${id_of_dep}.tree already in tree
394 otherlink=($(head -n 1 ${id_of_dep}.tree))
395 if [ -z "${otherlink[*]}" ]
396 then echo otherlink empty for $id_of_dep.tree
397 echo This should not happen, but happens to happen...
398 exit 1
399 fi
400#Do not use "${rootlink[*]}" =~ "${otherlink[*]}": case rootlink=(1 11)
401# and otherlink=(1 1)
402 if [[ ${srootlink#"${otherlink[*]} "} != ${srootlink} ]]; then # cir. dep
403 echo -en "\nCirc: $((depth+1))${spaceSTR:0:$((depth+1))}${YELLOW}${id_of_dep}${OFF} $dpriostring"
404# Find lowest priority in branch from parent to depFile:
405 p2=0
406 for (( i=${#otherlink[*]}; i < $depth ; i++ )) ; do
407 if (( ${priolink[i]} > $p2 )); then p2=${priolink[i]}; fi
408 done
409 if (( $prio_of_dep >= $p2 )); then # prune
410 lines_to_remove="$lines_to_remove $id_of_dep"
411 sed -i "/$id_of_dep/d" ${depFile/.tree/.dep}
412 else # find and set parent, then return lowest priority
413# The parent has the same link without the last entry.
414# We do not need otherlink anymore so just destroy the last element
415 unset otherlink[-1]
416 parentNode=$(grep ^"${otherlink[*]}"\$ -l *)
417 return $p2
418 fi
419 else # not circular: prune tree (but not .dep, since it may happen that
420 # the tree is destroyed and rebuilt in another order)
421 lines_to_remove="$lines_to_remove $id_of_dep"
422 fi # circular or not
423 continue # this dependency has already been seen, and the associated
424 # subtree computed. We are done
425 fi # Has this entry already been seen?
426# So, this entry has not already been seen. Let's build the corresponding
427# subtree. First check there is a subtree...
428# Use -s, because it may happen that after removing lines, .dep exists
429# but is empty.
430 if [[ -s ${id_of_dep}.dep ]]; then # this dependency has dependencies
431 sed "1i${rootlink[*]} $count\\
432${priolink[*]} $prio_of_dep" < ${id_of_dep}.dep \
433 > ${id_of_dep}.tree # add link and priolink
434 generate_dependency_tree ${id_of_dep}.tree $prio_of_dep
435# Test return value, in case we exchange dependencies
436 p2=$?
437 case $p2 in
438 0) # Normal return
439 ;;
440 $prio_of_dep) # we remove this dep, but since it may become unreachable,
441 # move it to be built later (as a dep of parent).
442 tree_erase ${id_of_dep}.tree
443 lines_to_remove="$lines_to_remove $id_of_dep"
444 sed -i "/${id_of_dep}/d" ${depFile/.tree/.dep}
445 echo "$prio_of_dep b $id_of_dep" >> $parentNode
446# must be added to .dep in case parentNode is destroyed when erasing
447# the tree
448 echo "$prio_of_dep b $id_of_dep" >> ${parentNode/.tree/.dep}
449 continue
450 ;;
451 *) # We are backing up
452 return $p2
453 ;;
454 esac
455 else # id_of_dep has no dependencies, just record the link in a file
456 # and print
457 echo "${rootlink[*]} $count" > ${id_of_dep}.tree
458 echo -en "\nLeaf: $(($depth+1))${spaceSTR:0:$(($depth+1))}${CYAN}${id_of_dep}${OFF} $dpriostring"
459 fi
460done
461echo -en "\n End: $depth${spaceSTR:0:$depth}${GREEN}${depFile%.tree}${OFF}"
462} <$depFile
463# It may happen that a file is created with several times
464# the same line. Normally, all those lines but one
465# would be flagged to be removed (or all of them if
466# the dependency appeared before). A simple sed /$line/d
467# destroys all the lines. We should instead remove
468# only one for each appearance of it in lines_to_remove.
469# so first get the position of last line and then delete
470# that line
471for line in $lines_to_remove
472 do lineno=$(sed -n /^[[:digit:]]\ b\ $line\$/= $depFile | tail -n1)
473 sed -i ${lineno}d $depFile
474done
475return 0
476}
477
478
479#---------------#
480tree_browse() { #
481#---------------#
482local file=$1
483local f
484
485#echo file=$file
486for f in $(grep '[^0-9 ]' $file | sed 's/.* //'); do
487# echo f=$f
488 if grep -q '[^0-9 ]' ${f}.tree ; then
489 tree_browse ${f}.tree
490 fi
491 echo $f
492done
493}
494
495#--------------#
496tree_erase() { #
497#--------------#
498local file=$1
499local f
500local rootlink
501local rootlink2
502
503#echo file=$file
504rootlink="$(head -n1 $file) "
505for f in $(grep '[^0-9 ]' $file | sed 's/.* //'); do
506 if [ -f ${f}.tree ]; then
507 rootlink2="$(head -n1 ${f}.tree) "
508# We want two things:
509# i) do not erase the file if it is in another branch
510# ii) do not erase the file if there is a circular dependency
511# for case i), we test that rootlink is contained in rootlink2
512# for case ii), we test that rootlink2 is not contained in
513# rootlink.
514# See comment above about srootlink
515 if [[ ${rootlink2#${rootlink}} != ${rootlink2} &&
516 ${rootlink#${rootlink2}} == ${rootlink} ]] ; then
517 tree_erase ${f}.tree
518 fi
519 fi
520done
521rm -f $file
522}
Note: See TracBrowser for help on using the repository browser.