my $TB = Test::More->builder;
if ($ENV{DBICTEST_IN_PERSISTENT_ENV}) {
- # without this explicit close ->reset below warns
- close ($TB->$_) for qw/output failure_output/;
+ # without this explicit close older TBs warn in END after a ->reset
+ if ($TB->VERSION < 1.005) {
+ close ($TB->$_) for (qw/output failure_output todo_output/);
+ }
- # so done_testing can work
- $TB->reset;
+ # if I do not do this, I get happy sigpipes on new TB, no idea why
+ # (the above close-and-forget doesn't work - new TB does *not* reopen
+ # its handles automatically anymore)
+ else {
+ for (qw/failure_output todo_output/) {
+ close $TB->$_;
+ open ($TB->$_, '>&', *STDERR);
+ }
+
+ close $TB->output;
+ open ($TB->output, '>&', *STDOUT);
+ }
- # this simulates a subtest
- $TB->_indent(' ' x 4);
+ # so done_testing can work on every persistent pass
+ $TB->reset;
}
use lib qw(t/lib);
use DBICTest::RunMode;
+use DBICTest::Util qw/populate_weakregistry assert_empty_weakregistry/;
use DBIx::Class;
use B 'svref_2object';
BEGIN {
if DBIx::Class::_ENV_::PEEPEENESS;
}
-use Scalar::Util qw/refaddr reftype weaken/;
-
# this is what holds all weakened refs to be checked for leakage
my $weak_registry = {};
# Skip the heavy-duty leak tracing when just doing an install
unless (DBICTest::RunMode->is_plain) {
- # have our own little stack maker - Carp infloops due to the bless override
- my $trace = sub {
- my $depth = 1;
- my (@stack, @frame);
-
- while (@frame = caller($depth++)) {
- push @stack, [@frame[3,1,2]];
- }
-
- $stack[0][0] = '';
- return join "\tinvoked as ", map { sprintf ("%s at %s line %d\n", @$_ ) } @stack;
- };
-
# redefine the bless override so that we can catch each and every object created
no warnings qw/redefine once/;
no strict qw/refs/;
}
);
- my $slot = (sprintf '%s=%s(0x%x)', # so we don't trigger stringification
- ref $obj,
- reftype $obj,
- refaddr $obj,
- );
+ # unicode is tricky, and now we happen to invoke it early via a
+ # regex in connection()
+ return $obj if (ref $obj) =~ /^utf8/;
- # weaken immediately to avoid weird side effects
- $weak_registry->{$slot} = { weakref => $obj, strace => $trace->() };
- weaken $weak_registry->{$slot}{weakref};
+ # Test Builder is now making a new object for every pass/fail (que bloat?)
+ # and as such we can't really store any of its objects (since it will
+ # re-populate the registry while checking it, ewwww!)
+ return $obj if (ref $obj) =~ /^TB2::/;
- return $obj;
+ # weaken immediately to avoid weird side effects
+ return populate_weakregistry ($weak_registry, $obj );
};
require Try::Tiny;
for my $func (qw/try catch finally/) {
my $orig = \&{"Try::Tiny::$func"};
*{"Try::Tiny::$func"} = sub (&;@) {
-
- my $slot = sprintf ('CODE(0x%x)', refaddr $_[0]);
-
- $weak_registry->{$slot} = { weakref => $_[0], strace => $trace->() };
- weaken $weak_registry->{$slot}{weakref};
-
+ populate_weakregistry( $weak_registry, $_[0] );
goto $orig;
}
}
# txn_do to invoke more codepaths
my ($mc_row_obj, $pager, $pager_explicit_count) = $schema->txn_do (sub {
- my $artist = $rs->create ({
+ my $artist = $schema->resultset('Artist')->create ({
name => 'foo artist',
cds => [{
title => 'foo cd',
return ($artist, $pg, $pg_wcount);
});
- # same for dbh_do
+ # more codepaths - error handling in txn_do
+ {
+ eval { $schema->txn_do ( sub {
+ $storage->_dbh->begin_work;
+ fail ('how did we get so far?!');
+ } ) };
+
+ eval { $schema->txn_do ( sub {
+ $schema->txn_do ( sub {
+ die "It's called EXCEPTION";
+ fail ('how did we get so far?!');
+ } );
+ fail ('how did we get so far?!');
+ } ) };
+ like( $@, qr/It\'s called EXCEPTION/, 'Exception correctly propagated in nested txn_do' );
+ }
+
+ # dbh_do codepath
my ($rs_bind_circref, $cond_rowobj) = $schema->storage->dbh_do ( sub {
my $row = $_[0]->schema->resultset('Artist')->new({});
my $rs = $_[0]->schema->resultset('Artist')->search({
}
}
- for (keys %$base_collection) {
- $weak_registry->{"basic $_"} = { weakref => $base_collection->{$_} };
- weaken $weak_registry->{"basic $_"}{weakref};
- }
+ populate_weakregistry ($weak_registry, $base_collection->{$_}, "basic $_")
+ for keys %$base_collection;
}
# check that "phantom-chaining" works - we never lose track of the original $schema
{
my $phantom;
for (
- sub { DBICTest->init_schema },
+ sub { DBICTest->init_schema( sqlite_use_file => 0 ) },
sub { shift->source('Artist') },
sub { shift->resultset },
sub { shift->result_source },
sub { shift->delete },
sub { shift->insert },
) {
- $phantom = $_->($phantom);
-
- my $slot = (sprintf 'phantom %s=%s(0x%x)', # so we don't trigger stringification
- ref $phantom,
- reftype $phantom,
- refaddr $phantom,
- );
-
- $weak_registry->{$slot} = $phantom;
- weaken $weak_registry->{$slot};
+ $phantom = populate_weakregistry ( $weak_registry, scalar $_->($phantom) );
}
ok( $phantom->in_storage, 'Properly deleted/reinserted' );
# T::B 2.0 has result objects and other fancyness
delete $weak_registry->{$slot};
}
+ elsif ($slot =~ /^Method::Generate::(?:Accessor|Constructor)/) {
+ # Moo keeps globals around, this is normal
+ delete $weak_registry->{$slot};
+ }
elsif ($slot =~ /^SQL::Translator/) {
# SQLT is a piece of shit, leaks all over
delete $weak_registry->{$slot};
delete $weak_registry->{$slot}
unless $cleared->{hash_merge_singleton}{$weak_registry->{$slot}{weakref}{behavior}}++;
}
+ elsif (
+ $slot =~ /^Data::Dumper/
+ and
+ $weak_registry->{$slot}{stacktrace} =~ /\QDBIx::Class::ResultSource::_mk_row_parser/
+ ) {
+ # there should be only one D::D object (used to construct the rowparser)
+ # more would indicate trouble
+ delete $weak_registry->{$slot}
+ unless $cleared->{mk_row_parser_dd_singleton}++;
+ }
elsif (DBIx::Class::_ENV_::INVISIBLE_DOLLAR_AT and $slot =~ /^__TxnScopeGuard__FIXUP__/) {
delete $weak_registry->{$slot}
}
or $r->result_source(undef);
}
-for my $slot (sort keys %$weak_registry) {
-
- ok (! defined $weak_registry->{$slot}{weakref}, "No leaks of $slot") or do {
- my $diag = '';
-
- $diag .= Devel::FindRef::track ($weak_registry->{$slot}{weakref}, 20) . "\n"
- if ( $ENV{TEST_VERBOSE} && eval { require Devel::FindRef });
-
- if (my $stack = $weak_registry->{$slot}{strace}) {
- $diag .= " Reference first seen$stack";
- }
-
- diag $diag if $diag;
- };
-}
+assert_empty_weakregistry ($weak_registry);
# we got so far without a failure - this is a good thing
# now let's try to rerun this script under a "persistent" environment
# this is ugly and dirty but we do not yet have a Test::Embedded or
# similar
-my @pperl_cmd = (qw/pperl --prefork=1/, __FILE__);
-my @pperl_term_cmd = @pperl_cmd;
-splice @pperl_term_cmd, 1, 0, '--kill';
+# set up -I
+require Config;
+$ENV{PERL5LIB} = join ($Config::Config{path_sep}, @INC);
+($ENV{PATH}) = $ENV{PATH} =~ /(.+)/;
+
+
+my $persistence_tests = {
+ PPerl => {
+ cmd => [qw/pperl --prefork=1/, __FILE__],
+ },
+ 'CGI::SpeedyCGI' => {
+ cmd => [qw/speedy -- -t5/, __FILE__],
+ },
+};
# scgi is smart and will auto-reap after -t amount of seconds
-my @scgi_cmd = (qw/speedy -- -t5/, __FILE__);
+# pperl needs an actual killer :(
+$persistence_tests->{PPerl}{termcmd} = [
+ $persistence_tests->{PPerl}{cmd}[0],
+ '--kill',
+ @{$persistence_tests->{PPerl}{cmd}}[ 1 .. $#{$persistence_tests->{PPerl}{cmd}} ],
+];
SKIP: {
skip 'Test already in a persistent loop', 1
if $ENV{DBICTEST_IN_PERSISTENT_ENV};
- skip 'Persistence test disabled on regular installs', 1
- if DBICTest::RunMode->is_plain;
-
skip 'Main test failed - skipping persistent env tests', 1
unless $TB->is_passing;
- # set up -I
- require Config;
- local $ENV{PERL5LIB} = join ($Config::Config{path_sep}, @INC);
-
local $ENV{DBICTEST_IN_PERSISTENT_ENV} = 1;
- # try with pperl
- SKIP: {
- skip 'PPerl persistent environment tests require PPerl', 1
- unless eval { require PPerl };
+ require IPC::Open2;
+
+ for my $type (keys %$persistence_tests) { SKIP: {
+ skip "$type module not found", 1
+ unless eval "require $type";
+
+ my @cmd = @{$persistence_tests->{$type}{cmd}};
# since PPerl is racy and sucks - just prime the "server"
{
local $ENV{DBICTEST_PERSISTENT_ENV_BAIL_EARLY} = 1;
- system(@pperl_cmd);
+ system(@cmd);
sleep 1;
- # see if it actually runs - if not might as well bail now
- skip "Something is wrong with pperl ($!)", 1
- if system(@pperl_cmd);
+ # see if the thing actually runs, if not - might as well bail now
+ skip "Something is wrong with $type ($!)", 1
+ if system(@cmd);
}
for (1,2,3) {
- system(@pperl_cmd);
- ok (!$?, "Run in persistent env (PPerl pass $_): exit $?");
- }
-
- ok (! system (@pperl_term_cmd), 'killed pperl instance');
- }
-
- # try with speedy-cgi
- SKIP: {
- skip 'SPeedyCGI persistent environment tests require CGI::SpeedyCGI', 1
- unless eval { require CGI::SpeedyCGI };
-
- {
- local $ENV{DBICTEST_PERSISTENT_ENV_BAIL_EARLY} = 1;
- skip "Something is wrong with speedy ($!)", 1
- if system(@scgi_cmd);
- sleep 1;
+ note ("Starting run in persistent env ($type pass $_)");
+ IPC::Open2::open2(my $out, undef, @cmd);
+ my @out_lines;
+ while (my $ln = <$out>) {
+ next if $ln =~ /^\s*$/;
+ push @out_lines, " $ln";
+ last if $ln =~ /^\d+\.\.\d+$/; # this is persistence, we need to terminate reading on our end
+ }
+ print $_ for @out_lines;
+ close $out;
+ wait;
+ ok (!$?, "Run in persistent env ($type pass $_): exit $?");
+ ok (scalar @out_lines, "Run in persistent env ($type pass $_): got output");
}
- for (1,2,3) {
- system(@scgi_cmd);
- ok (!$?, "Run in persistent env (SpeedyCGI pass $_): exit $?");
- }
- }
+ ok (! system (@{$persistence_tests->{$type}{termcmd}}), "killed $type server instance")
+ if $persistence_tests->{$type}{termcmd};
+ }}
}
done_testing;
# PID files to go by (man does pperl really suck :(
END {
unless ($ENV{DBICTEST_IN_PERSISTENT_ENV}) {
- close STDOUT;
- close STDERR;
+ close $_ for (*STDIN, *STDOUT, *STDERR);
local $?; # otherwise test will inherit $? of the system()
- system (@pperl_term_cmd);
+ system (@{$persistence_tests->{PPerl}{termcmd}});
}
}
projects / dbsrgits/DBIx-Class.git / blobdiff