avoid removing empty list of packages

[lexming]

Fixes #1553

disclaimer: I'm not an expert in Perl, so better solutions may exist.

[stdweird]

i would do here

(defined($to_rm) && $to_rm) or return 0;

() is for readability; and untested ofcourse ;)

[lexming]

An empty $to_rm does not evaluate to false, so (defined($to_rm) && $to_rm) or return 0; continues execution.

[lexming]

Probably because an empty $to_rm is () and that is something evaluated as a string (just a guess)

[stdweird]

weird, i would have expected an empty Set::Scalar looking at the code. packages_to_remove shouldn't return an undef, so i'm probably missng something here why there is a defined in the first place...

[ned21]

I think the defined($to_rm) is intended as error handling in the specific case that the function returned undef, there's no reason for the rest of the code not to run if there is nothing to remove?

[jrha]

Looking at this very briefly, it looks like for some reason $to_rm is not a Set::Scalar here and it should be.

[lexming]

I found the actual underlying cause of this issue in our system. The output of the LEAF_PACKAGES command contains some extra warning messages at the top with blank lines:

$ package-cleanup -c /etc/yum.conf --leaves --all --qf "%{NAME};%{ARCH}"
Loaded plugins: priorities, product-id, subscription-manager, versionlock

This system is not registered with an entitlement server. You can use subscription-manager to register.

SINDES-ca-certificate-cerberus;noarch
SOAPpy;noarch
[...]

The regex in packages_to_remove() does not handle blank lines:

configuration-modules-core/ncm-spma/src/main/perl/spma/yum.pm

Line 697 in bdb85e6

my $leaves = Set::Scalar->new(grep($_ !~ m{\s}, split(/\n/, $out)));

Each blank line becomes an element of $to_rm with an empty string in it.

Fix for the regex in f0c309b

First change has been made, but I have a new comment

[ned21]

Thanks for finding the root cause, @lexming ! I think it's still worth leaving in the length check in addition the regex fix, but would you mind using git rebase -i to squash b537693 into 45a4cc6? That will keep the history "clean" without confusing things with the previous version was never merged.

[lexming]

@ned21 sure, commits squashed 👍

[jrha]

Looks like the tests are failing on:

2022-09-13T21:09:56.2659621Z Can't use string ("1") as an ARRAY ref while "strict refs" in use at /__w/configuration-modules-core/configuration-modules-core/ncm-spma/target/lib/perl/NCM/Component/spma/yum.pm line 922.
2022-09-13T21:09:56.2666127Z # Tests were run but no plan was declared and done_testing() was not seen.
2022-09-13T21:09:56.2735212Z src/test/perl/yum-update-pkgs.t .................