Error during registry setup steps when deploying from Windows

[sbuttgereit]

When performing an initial deployment from a Windows 10 system, the following error is encountered:

> sqitch deploy db:pg://scb@127.0.0.1:5432/fit_data
Adding registry tables to db:pg://scb@127.0.0.1:5432/fit_data
Use of uninitialized value $ret in scalar chomp at C:/strawberry/perl/site/lib/App/Sqitch.pm line 484.

For the record this is Windows 10 using Strawberry Perl (This is perl 5, version 20, subversion 1 (v5.20.1) built for MSWin32-x64-multi-thread) installed via Chocolatey; Sqitch is version .9994 installed via cpanminus.

I know nothing of Perl (this is really the first Perl program I've tried to comprehend), but I think I've at least traced this issue to where it's happening. The error is thrown when sqitch.pm is trying to figure out whether we're on Postgres XC or not... specifically (starting line 189):

# Is this XC?
    $opts = ' DISTRIBUTE BY REPLICATION' if $self->_probe('-c', q{
        SELECT count(*)
          FROM pg_catalog.pg_proc p
          JOIN pg_catalog.pg_namespace n ON p.pronamespace = n.oid
         WHERE nspname = 'pg_catalog'
           AND proname = 'pgxc_version';
    });

Ultimately this ends up in a "probe" subroutine which in turn calls "capture". Adding print statements, through these calls appear (at least to my uneducated eye) to be sensible... the command line parameters generally. I get to this routine in subroutine capture in sqitch.pm (line 362):

sub capture {
    my $self = shift;
    local $SIG{__DIE__} = sub {
        ( my $msg = shift ) =~ s/\s+at\s+.+/\n/ms;
        die $msg;
    };
    capturex @_;
}

Up to the capturex line, my print statements seem sensible. I assume that the return of capturex is the return of the capture subroutine (which gets passed back to subroutine probe where the error is actually thrown.) I know that capturex is from another package that Sqitch depends on, so this may not ultimately be a Sqitch problem... but reporting here as it is ultimately a blocker for anyone in a pure Windows environment.

I have tested my setup and project using a Linux environment and things are fine there (same version of Sqitch at least). I also have tested running Sqitch commands in Windows after the Linux environment has built out the registry and every else seems to work fine in Windows. Finally I have also tested in the new Windows Subsystem for Linux... not even close getting Sqitch to install into that environment so, but that's not a Sqitch problem as yet (and for some while I expect).

[theory]

Thank you. Does this patch fix the issue?

diff --git a/lib/App/Sqitch.pm b/lib/App/Sqitch.pm
index 4f59a54..dd6da36 100644
--- a/lib/App/Sqitch.pm
+++ b/lib/App/Sqitch.pm
@@ -481,7 +481,7 @@ sub spool {

 sub probe {
     my ($ret) = shift->capture(@_);
-    chomp $ret;
+    chomp $ret if $ret;
     return $ret;
 }

[sbuttgereit]

Sorry for not responding sooner. Just tested and things seem to work as expected with that change. I was able to successfully deploy from Windows to a freshly created database with no registry. The registry was created as expected and I see the expected data in the registry.

[theory]

Awesome, thanks!

[decibel]

FWIW, EXISTS() would be more efficient than count(). Probably a non-issue in this case, but it's a better pattern to follow. It also makes it clearer that we only care if there's a single row, not about an exact count.

[theory]

Meh, should only ever return 1 or 0.