From: Peter Rabbitson Date: Tue, 4 Mar 2014 08:39:49 +0000 (+0100) Subject: Fix pesky on_connect_* race condition abraxxa++ ilmari++ X-Git-Tag: v0.082800~232 X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=commitdiff_plain;h=1eb87dd767c4bdb815085acb2a8e63e12b32f990;p=dbsrgits%2FDBIx-Class.git Fix pesky on_connect_* race condition abraxxa++ ilmari++ A race condition existed between storage accessor setters and the determine_driver routines, triggering a connection before the set-cycle is finished --- diff --git a/Changes b/Changes index d557c39..387cb0a 100644 --- a/Changes +++ b/Changes @@ -1,5 +1,10 @@ Revision history for DBIx::Class + * Fixes + - Fix on_connect_* not always firing in some cases - a race condition + existed between storage accessor setters and the determine_driver + routines, triggering a connection before the set-cycle is finished + 0.08270 2014-01-30 21:54 (PST) * Fixes - Fix 0.08260 regression in DBD::SQLite bound int handling. Inserted diff --git a/lib/DBIx/Class/Storage/DBI.pm b/lib/DBIx/Class/Storage/DBI.pm index 23a7f71..9b5e3b0 100644 --- a/lib/DBIx/Class/Storage/DBI.pm +++ b/lib/DBIx/Class/Storage/DBI.pm @@ -79,20 +79,24 @@ __PACKAGE__->_use_join_optimizer (1); sub _determine_supports_join_optimizer { 1 }; # Each of these methods need _determine_driver called before itself -# in order to function reliably. This is a purely DRY optimization +# in order to function reliably. We also need to separate accessors +# from plain old method calls, since an accessor called as a setter +# does *not* need the driver determination loop fired (and in fact +# can produce hard to find bugs, like e.g. losing on_connect_* +# semantics on fresh connections) # -# get_(use)_dbms_capability need to be called on the correct Storage -# class, as _use_X may be hardcoded class-wide, and _supports_X calls -# _determine_supports_X which obv. needs a correct driver as well -my @rdbms_specific_methods = qw/ +# The construct below is simply a parameterized around() +my $storage_accessor_idx = { map { $_ => 1 } qw( sqlt_type - deployment_statements + datetime_parser_type sql_maker cursor_class +)}; +for my $meth (keys %$storage_accessor_idx, qw( + deployment_statements build_datetime_parser - datetime_parser_type txn_begin @@ -110,15 +114,13 @@ my @rdbms_specific_methods = qw/ _server_info _get_server_version -/; - -for my $meth (@rdbms_specific_methods) { +)) { my $orig = __PACKAGE__->can ($meth) or die "$meth is not a ::Storage::DBI method!"; - no strict qw/refs/; - no warnings qw/redefine/; + no strict 'refs'; + no warnings 'redefine'; *{__PACKAGE__ ."::$meth"} = subname $meth => sub { if ( # only fire when invoked on an instance, a valid class-based invocation @@ -129,6 +131,10 @@ for my $meth (@rdbms_specific_methods) { and ! $_[0]->{_in_determine_driver} and + # if this is a known *setter* - just set it, no need to connect + # and determine the driver + ! ( $storage_accessor_idx->{$meth} and @_ > 1 ) + and # Only try to determine stuff if we have *something* that either is or can # provide a DSN. Allows for bare $schema's generated with a plain ->connect() # to still be marginally useful diff --git a/t/storage/on_connect_do.t b/t/storage/on_connect_do.t index 2874a9d..115fadb 100644 --- a/t/storage/on_connect_do.t +++ b/t/storage/on_connect_do.t @@ -36,6 +36,11 @@ $schema->storage->disconnect; ok $schema->connection( sub { DBI->connect(DBICTest->_database, undef, undef, { AutoCommit => 0 }) }, { + # DO NOT REMOVE - this seems like an unrelated piece of info, + # but is in fact a test for a bug where setting an accessor-via-option + # would trigger an early connect *bypassing* the on_connect_* pieces + cursor_class => 'DBIx::Class::Storage::Cursor', + on_connect_do => [ 'CREATE TABLE TEST_empty (id INTEGER)', [ 'INSERT INTO TEST_empty VALUES (?)', {}, 2 ],