refactor

Makefile

@@ -1,14 +1,15 @@
-testinput: run.py
-	./run.py /var/lib/apt/lists/ftp.debian.org_debian_dists_sid_main_binary-amd64_Packages > testinput
+.PHONY: all
+all: versions.rand versions.sorted testdpkg testapt testdose
+	set -e; for t in testapt.py testdpkg.pl testdpkg testapt testdose; do \
+		printf "running $$t..." >&2; \
+		./$$t < versions.rand | cmp versions.sorted - && echo OK >&2; \
+	done
 
-testinput.1000: run.py
-	./run.py /var/lib/apt/lists/ftp.debian.org_debian_dists_sid_main_binary-amd64_Packages | awk 'NR % 1000 == 1' > testinput.1000
-
-testinput.unequal: testinput inequalityfilter
-	./inequalityfilter < testinput > testinput.unequal
-
-testinput.unequal.1000: testinput inequalityfilter
-	./inequalityfilter < testinput | awk 'NR % 1000 == 1' > testinput.unequal.1000
+versions.rand: versions.sorted
+	# create a randomized version list using a reproducible seed value
+	{ openssl enc -aes-256-ctr -pass pass:1612543740 -nosalt </dev/zero 2>/dev/null 4<&- | { \
+	  tr ' ' '\n' <versions.sorted 3<&- | sort --random-source=/proc/self/fd/3 --random-sort; \
+	} 3<&0 <&4 4<&-; } 4<&0 > versions.rand
 
 testdpkg: testdpkg.c
 	gcc testdpkg.c -ldpkg -o testdpkg
@@ -16,34 +17,25 @@ testdpkg: testdpkg.c
 testapt: testapt.cc
 	g++ testapt.cc -lapt-pkg -o testapt
 
-testdose: testdose.ml
-	ocamlfind ocamlc -package dose3 -linkpkg testdose.ml -o testdose
+testdose:
+	ocamlfind ocamlc -package dose3.versioning -linkpkg testdose.ml -o testdose
 
-inequalityfilter: inequalityfilter.cc
-	g++ inequalityfilter.cc -lapt-pkg -o inequalityfilter
-
-.PHONY: test
-test: testinput testdpkg testapt testdose rundpkg.sh
-	./rundpkg.sh < testinput | md5sum
-	./testapt < testinput | md5sum
-	./testdose < testinput | md5sum
-
-.PHONY: test.1000
-test.1000: testinput.1000 testdpkg testapt testdose rundpkg.sh
-	./rundpkg.sh < testinput.1000 | md5sum
-	./testapt < testinput.1000 | md5sum
-	./testdose < testinput.1000 | md5sum
-
-.PHONY: testsort
-testsort: testinput.unequal testapt testsort.sh
-	./testapt < testinput.unequal | md5sum
-	./testsort.sh < testinput.unequal | md5sum
-
-.PHONY: testsort.1000
-testsort.1000: testinput.unequal.1000 testapt testsort.sh
-	./testapt < testinput.unequal.1000 | md5sum
-	./testsort.sh < testinput.unequal.1000 | md5sum
+versions.sorted:
+	wget https://metasnap.debian.net/by-package/debian.sqlite3
+	wget https://metasnap.debian.net/by-package/debian-backports.sqlite3
+	wget https://metasnap.debian.net/by-package/debian-debug.sqlite3
+	wget https://metasnap.debian.net/by-package/debian-ports.sqlite3
+	wget https://metasnap.debian.net/by-package/debian-security.sqlite3
+	wget https://metasnap.debian.net/by-package/debian-volatile.sqlite3
+	{ sqlite3 debian.sqlite3 "select name from vers"; \
+	  sqlite3 debian-backports.sqlite3 "select name from vers"; \
+	  sqlite3 debian-debug.sqlite3 "select name from vers"; \
+	  sqlite3 debian-ports.sqlite3 "select name from vers"; \
+	  sqlite3 debian-security.sqlite3 "select name from vers"; \
+	  sqlite3 debian-volatile.sqlite3 "select name from vers"; \
+	} | sort -u | grep '^[0-9]' | ./testapt.py > versions.sorted
+	rm debian.sqlite3 debian-backports.sqlite3 debian-debug.sqlite3 debian-ports.sqlite3 debian-security.sqlite3 debian-volatile.sqlite3
 
 .PHONY: clean
 clean:
-	rm -f testapt testdose testdose.cmi testdose.cmo testdpkg testinput inequalityfilter
+	rm -f testapt testdose testdose.cmi testdose.cmo testdpkg versions.rand

README.md

@@ -1,18 +1,23 @@
-This is to test the several implementations of Debian version comparison
-algorithms.
+Comparing Debian versions correctly is hard. This project allows comparing
+existing implementation by sorting a text file containing all Debian versions
+that have been used since 2005.
 
-Run it via `make test`.
+To run it, make sure you have the following packages installed:
 
-`run.py` generates a file containing all permutations of pairs of versions. The
-different programs then check whether versions are greater than, equal or less
-than one another and output >, = and <, respectively in order. It is then
-checked whether the output of all implementations is the same.
+    $ sudo apt install build-essenital libdpkg-dev libapt-pkg-dev libextlib-ocaml-dev ocaml-findlib ocaml libdose3-ocaml-dev
 
-Currently, this checks the libdpkg C implementation, libapt and dose3.
+Then execute:
+
+    $ make
+
+The following parsers are currently tested:
+
+ - dpkg (C and Perl)
+ - apt (C++ and Python)
+ - dose3 (OCaml)
 
 Additionally, the following parsers could be added:
 
- - libdpkg-perl
  - libben-ocaml (in lib/benl_base.ml and lib/benl_dpkg.c)
 
 The following do not use their own parser:
@@ -21,3 +26,9 @@ The following do not use their own parser:
  - libghc-dpkg-dev uses libdpkg-dev
  - ruby-debian uses libapt-pkg-dev
  - dpkg-dev-el uses `dpkg --compare-versions`
+
+Each of the scripts parses a list of versions coming from standard input, one
+per line, and writes them in sorted ascending order to standard output, one per
+line. Versions that compare as equal are written on the same line,
+lexicographically sorted and separated by space. The ground truth output is
+stored in the file `versions.sorted`.

inequalityfilter.cc

@@ -1,41 +0,0 @@
-#include <apt-pkg/debversion.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-
-int main()
-{
-    char *line = NULL;
-    char *token;
-    char *orig1;
-    char *orig2;
-    char *ver1;
-    char *ver2;
-    size_t len = 0;
-    size_t read;
-    int ret;
-    while ((read = getline(&line, &len, stdin)) != -1) {
-        //fprintf(stderr, "%s", line);
-        orig1 = strdup(line);
-        orig2 = strdup(line);
-        token = orig1;
-        ver1 = strsep(&token, "\t");
-        if (ver1 == NULL) {
-            fprintf(stderr, "cannot read token1");
-            exit(EXIT_FAILURE);
-        }
-        ver2 = strsep(&token, "\n");
-        if (ver2 == NULL) {
-            fprintf(stderr, "cannot read token2");
-            exit(EXIT_FAILURE);
-        }
-        ret = debVS.CmpVersion(ver1, ver2);
-        if (ret != 0) {
-            fputs(orig2, stdout);
-        }
-        //fputs(orig2, stdout);
-        free(orig1);
-        free(orig2);
-    }
-    exit(EXIT_SUCCESS);
-}

run.py

@@ -1,22 +0,0 @@
-#!/usr/bin/env python
-
-from __future__ import print_function
-import deb822
-import sys
-
-versions = set()
-with open(sys.argv[1]) as f:
-    for pkg in deb822.Deb822.iter_paragraphs(f):
-        ver = pkg.get('Version')
-        if ver:
-            versions.add(ver)
-
-# create a list of all possible combinations of two versions by iterating
-# through the list of versions and comparing every version with all that come
-# after it
-versions = sorted(versions)
-l = len(versions)
-for i,v1 in enumerate(versions):
-    print("%f %%\r"%((i*100.0)/l), file=sys.stderr, end="")
-    for v2 in versions[i+1:]:
-        print("%s\t%s"%(v1,v2))

rundpkg.sh

@@ -1,3 +0,0 @@
-#!/bin/sh
-# we need to do this because of bug#47214
-split --lines=1000000 --numeric-suffixes --suffix-length=4 --filter='./testdpkg'

testapt.cc

@@ -3,47 +3,44 @@
 #include <stdlib.h>
 #include <string.h>
 
+int cmpver(const void *a, const void *b) {
+	int ret = debVS.CmpVersion(*(const char **) a, *(const char **) b);
+	if (ret == 0) {
+		return strcmp(*(const char **) a, *(const char **) b);
+	}
+	return ret;
+}
+
+
 int main()
 {
     char *line = NULL;
-    char *token;
-    char *orig1;
-    char *orig2;
-    char *ver1;
-    char *ver2;
     size_t len = 0;
     size_t read;
-    int ret;
+	char **versions = NULL;
+	size_t num_versions = 0;
     while ((read = getline(&line, &len, stdin)) != -1) {
-        //fprintf(stderr, "%s", line);
-        orig1 = strdup(line);
-        orig2 = strdup(line);
-        token = orig1;
-        ver1 = strsep(&token, "\t");
-        if (ver1 == NULL) {
-            fprintf(stderr, "cannot read token1");
-            exit(EXIT_FAILURE);
-        }
-        ver2 = strsep(&token, "\n");
-        if (ver2 == NULL) {
-            fprintf(stderr, "cannot read token2");
-            exit(EXIT_FAILURE);
-        }
-        ret = debVS.CmpVersion(ver1, ver2);
-        if (ret == 0) {
-            fputc('=', stdout);
-        } else if (ret > 0) {
-            fputc('>', stdout);
-        } else {
-            fputc('<', stdout);
-        }
-        /*
-        fputc(' ', stdout);
-        fputs(orig2, stdout);
-        */
-        //fputs(orig2, stdout);
-        free(orig1);
-        free(orig2);
-    }
+		num_versions += 1;
+		versions = (char **)realloc(versions, sizeof(char *)*num_versions);
+		if (versions == NULL) {
+			perror("malloc failed");
+			exit(EXIT_FAILURE);
+		}
+		if (line[read-1] == '\n') {
+			line[read-1] = '\0';
+		}
+		versions[num_versions-1] = strdup(line);
+	}
+	qsort(versions, num_versions, sizeof(char *), cmpver);
+	printf(versions[0]);
+	for (int i = 1; i < num_versions; i++) {
+		if (debVS.CmpVersion(versions[i-1], versions[i]) == 0) {
+			printf(" ");
+		} else {
+			printf("\n");
+		}
+		printf(versions[i]);
+	}
+	printf("\n");
     exit(EXIT_SUCCESS);
 }

testapt.py

@@ -0,0 +1,29 @@
+#!/usr/bin/env python3
+
+import sys
+from functools import cmp_to_key
+import apt_pkg
+
+apt_pkg.init_system()
+
+
+def myvercomp(a, b):
+    ret = apt_pkg.version_compare(a, b)
+    if ret == 0:
+        return 1 if a > b else -1
+    return ret
+
+
+l = [line.rstrip("\n") for line in sys.stdin]
+l = sorted(l, key=cmp_to_key(myvercomp))
+
+last = l.pop(0)
+print(last, end="")
+for e in l:
+    if apt_pkg.version_compare(last, e) == 0:
+        print(" ", end="")
+    else:
+        print()
+    last = e
+    print(e, end="")
+print()

testdose.ml

@@ -1,12 +1,17 @@
-open Debian
 open ExtLib
 
 let main () =
-  List.iter (fun line ->
-      let ver1, ver2 = String.split line "\t" in
-      let ret = Version.compare ver1 ver2 in
-      print_char (if ret > 0 then '>' else if ret < 0 then '<' else '=')
-    ) (Std.input_list stdin);
+  let versions = List.sort ~cmp:(fun a b ->
+      let ret = Dose_versioning.Debian.compare a b in
+      if ret == 0 then compare a b
+      else ret
+    ) (Std.input_list stdin) in
+  print_string (List.hd versions);
+  ignore (List.fold_left (fun acc e ->
+          if Dose_versioning.Debian.compare acc e == 0 then print_char ' '
+          else print_char '\n'; print_string e; e
+    ) (List.hd versions) (List.tl versions));
+  print_char '\n';
 ;;
 
 main ();;

testdpkg.c

@@ -1,57 +1,62 @@
 #define LIBDPKG_VOLATILE_API 1
 #define _GNU_SOURCE
 #include <dpkg/dpkg.h>
+#include <dpkg/dpkg-db.h>
 #include <dpkg/version.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <errno.h>
+
+// we need to keep track of the original string because some version strings
+// include the implicit epoch of zero
+struct myver {
+	struct dpkg_version dv;
+	char *sv;
+};
+
+int cmpver(const void *a, const void *b) {
+	int ret = dpkg_version_compare(&(((struct myver *)a)->dv), &(((struct myver *)b)->dv));
+	if (ret == 0) {
+		return strcmp(((struct myver *)a)->sv, ((struct myver *)b)->sv);
+	}
+	return ret;
+}
 
 int main()
 {
     char *line = NULL;
-    char *token;
-    char *orig1;
-    char *orig2;
-    char *ver1;
-    char *ver2;
     size_t len = 0;
     size_t read;
-    struct dpkg_version a, b;
-    int ret;
+	struct dpkg_version a;
+	struct myver *versions = NULL;
+	size_t num_versions = 0;
     while ((read = getline(&line, &len, stdin)) != -1) {
-        //fprintf(stderr, "%s", line);
-        orig1 = strdup(line);
-        orig2 = strdup(line);
-        token = orig1;
-        ver1 = strsep(&token, "\t");
-        if (ver1 == NULL) {
-            fprintf(stderr, "cannot read token1");
-            exit(EXIT_FAILURE);
-        }
-        if(parseversion(&a, ver1, NULL)) {
-            fprintf(stderr, "cannot parse version1 %s\n", ver1);
-            exit(EXIT_FAILURE);
-        }
-        ver2 = strsep(&token, "\n");
-        if (ver2 == NULL) {
-            fprintf(stderr, "cannot read token2");
-            exit(EXIT_FAILURE);
-        }
-        if(parseversion(&b, ver2, NULL)) {
-            fprintf(stderr, "cannot parse version2 %s\n", ver2);
-            exit(EXIT_FAILURE);
-        }
-        ret = dpkg_version_compare(&a, &b);
-        if (ret == 0) {
-            fputc('=', stdout);
-        } else if (ret > 0) {
-            fputc('>', stdout);
-        } else {
-            fputc('<', stdout);
-        }
-        //fputs(orig2, stdout);
-        free(orig1);
-        free(orig2);
-    }
+		num_versions += 1;
+		versions = realloc(versions, sizeof(struct myver)*num_versions);
+		if (versions == NULL) {
+			perror("malloc failed");
+			exit(EXIT_FAILURE);
+		}
+		if (line[read-1] == '\n') {
+			line[read-1] = '\0';
+		}
+		if (parseversion(&((versions+num_versions-1)->dv), line, NULL)) {
+			fprintf(stderr, "cannot parse version: %s\n", line);
+			exit(EXIT_FAILURE);
+		}
+		(versions+num_versions-1)->sv = strdup(line);
+	}
+	qsort(versions, num_versions, sizeof(struct myver), cmpver);
+	printf(versions->sv);
+	for (int i = 1; i < num_versions; i++) {
+		if (dpkg_version_compare(&((versions+i-1)->dv), &((versions+i)->dv)) == 0) {
+			printf(" ");
+		} else {
+			printf("\n");
+		}
+		printf((versions+i)->sv);
+	}
+	printf("\n");
     exit(EXIT_SUCCESS);
 }

testdpkg.pl

@@ -0,0 +1,41 @@
+#!/usr/bin/perl
+
+use strict;
+use warnings;
+
+use Dpkg::Version;
+
+my @versions = ();
+while ( my $line = <> ) {
+    chomp $line;
+    my $ver = Dpkg::Version->new( $line, check => 1 );
+    die "not a valid version: $line" if !defined $ver;
+    push @versions, [ $ver, $line ];
+}
+
+sub myvercomp {
+
+    # Dpkg::Version overloads <=>
+    my $ret = $a->[0] <=> $b->[0];
+    if ( $ret == 0 ) {
+
+        # fall back to string comparison
+        return $a->[1] cmp $b->[1];
+    }
+    return $ret;
+}
+@versions = sort myvercomp @versions;
+
+my $last = shift @versions;
+print "$last->[1]";
+foreach my $v (@versions) {
+    if ( ( $last->[0] <=> $v->[0] ) == 0 ) {
+        print " ";
+    }
+    else {
+        print "\n";
+    }
+    $last = $v;
+    print "$v->[1]";
+}
+print "\n";

testsort.sh

@@ -1,29 +0,0 @@
-#!/bin/sh
-
-# some examples where `sort -V` does the wrong thing:
-# 0.04-1-5             0.4-3
-# 0.1+20080921-2       0.1-1+b8
-# 0.1+dfsg-4           000.001-4
-# 0.1.0+20071012-1.2   0.1.0-1.1
-# 0.3.0+20091229-1     0.3.0-7.2
-# 0.6.0+git20130305-5  0.6.0-5+b1
-# 0:2009.10.04-1       1.0pre11-1
-# 0:2009.10.04-1       1.240-1
-# 0:2009.10.04-1       1.7.2.4-4.1
-# 0:2009.10.04-1       2.3-12
-# 0:2009.10.04-1       8.0.184.15484+dfsg-2
-# 1.0+dfsg-1           1.00-6
-
-i=0
-while read line; do
-	printf "$i\r" >&2
-	i=$((i+1))
-	set -- $line
-	newest=$( ( echo "$1"; echo "$2" ) | sort -V | tail -n1)
-	if [ "$1" != "$newest" ]; then
-		printf "<"
-	else
-		printf ">"
-	fi
-	printf " %s\t%s\n" $1 $2
-done