sablink/devkit
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Cache results of package_deps and re-enable virtual dep stripping

Browse Source 
It's likely that some virtuals are dependencies of multiple package arguments
to builder and the depgraph would be queried multiple times. This commit
caches the results to avoid expensive duplicate shell execs and optimise for
speed.

Additionally, when calls to equery depgraph match multiple packages (e.g.
the query is for an unversioned atom, and multiple versions exist in the
portage tree), the results for all versions are returned, resulting in
duplicates. This change also strips out the duplicate entries so further
work need only be done once.

This uses perl 5.10's `state` variables, so adds the appropriate use. Since
Sabayon ships only perl 5.20, this should not be an issue.
This commit is contained in:
Ben Roberts
2016-03-20 19:43:20 +00:00
parent 566e4603c8
commit 967cd9708a
1 changed files with 38 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files
builder
+38 -16
View File
@@ -1,6 +1,8 @@
#!/usr/bin/env perl
use Getopt::Long;
use List::MoreUtils qw(uniq);
use v5.10;
my $profile = $ENV{BUILDER_PROFILE} // 3;
my $jobs = $ENV{BUILDER_JOBS} // 1;
@@ -42,9 +44,30 @@ sub package_deps {
my $package = shift;
my $depth = shift // 1; # defaults to 1 level of depthness of the tree
my $atom = shift // 0;
return
map { $_ =~ s/\[.*\]|\s//g; &atom($_) if $atom; $_ }
qx/equery -C -q g --depth=$depth $package/; #depth=0 it's all
# Since we expect this sub to be called multiple times with the same arguments, cache the results
state %cache;
$cache_key = "${package}:${depth}:${atom}";
if ( ! exists $cache{$cache_key} ) {
@dependencies = qx/equery -C -q g --depth=$depth $package/; #depth=0 it's all
chomp @dependencies;
# If an unversioned atom is given, equery returns results for all versions in the portage tree
# leading to duplicates. The sanest thing to do is dedup the list. This gives the superset of all
# possible dependencies, which isn't perfectly accurate but should be good enough. For completely
# accurate results, pass in a versioned atom.
@dependencies =
uniq
sort
grep { $_ }
map { $_ =~ s/\[.*\]|\s//g; &atom($_) if $atom; $_ }
@dependencies;
$cache{$cache_key} = \@dependencies;
}
return @{ $cache{$cache_key} };
}
#Input: nothing
@@ -71,18 +94,17 @@ sub calculate_missing {
say "[$package] Getting the package dependencies and the installed packages";
my @dependencies = package_deps( $package, $depth, 1 );
# XXX: Endless loop as for now.
# my %install_dependencies = map { $_ => 1 } @dependencies;
# # Look for any virtuals and remove its immediate dependencies to avoid
# # installing multiple conflicting packages one by one
# my @virtual_deps;
# for my $dep (@dependencies) {
# push(@virtual_deps, package_deps( $dep, 1, 1 )) if ( $dep =~ /^virtual\// );
# }
# for my $dep (@virtual_deps) {
# $install_dependencies{$dep} = 0 if ( $dep !~ /^virtual\// );
# }
# @dependencies = grep { $install_dependencies{$_} } @dependencies;
my %install_dependencies = map { $_ => 1 } @dependencies;
# Look for any virtuals and remove its immediate dependencies to avoid
# installing multiple conflicting packages one by one
my @virtual_deps;
for my $dep (@dependencies) {
push(@virtual_deps, package_deps( $dep, 1, 1 )) if ( $dep =~ /^virtual\// );
}
for my $dep (@virtual_deps) {
$install_dependencies{$dep} = 0 if ( $dep !~ /^virtual\// );
}
@dependencies = grep { $install_dependencies{$_} } @dependencies;
#taking only the 4th column of output as key of the hashmap
my %installed_packs =
@@ -93,7 +115,7 @@ sub calculate_missing {
my @to_install = grep( defined $available_packs{$_},
uniq( grep( !defined $installed_packs{$_}, @dependencies ) ) );
@to_install=grep { length } @to_install;
# say "[$package] packages that will be installed with equo: @to_install" if @to_install>0;
say "[$package] packages that will be installed with equo: @to_install" if @to_install>0;
return @to_install;
}