solver: non preferred actions are worse then non preferred pinning

Otherwise we might start to change packages unexpectedly when not
upgrading. This also fixes some other things the solver might've
decided to do.

Add also few test cases to detect bad behaviour.

3 files changed, 39 insertions, 0 deletions

diff --git a/test/pinning.installed2 b/test/pinning.installed2
new file mode 100644
index 0000000..c41ddf7
--- /dev/null
+++ b/test/pinning.installed2
@@ -0,0 +1,20 @@
+C:Q1eVpkasfqZAukAXFYbgwt4xffZWU=
+P:a
+V:3
+S:1
+I:1
+D:b
+
+C:Q1hdUpqRv5mYgJEqW52UmVsv23ysE=
+P:b
+V:3
+S:1
+I:1
+
+C:Q1eVpkasfqZAukAXFYbg324xAt4WU=
+P:c
+V:3
+S:1
+I:1
+D:a>=3
+
diff --git a/test/pinning8.test b/test/pinning8.test
new file mode 100644
index 0000000..84f312b
--- /dev/null
+++ b/test/pinning8.test
@@ -0,0 +1,11 @@
+@ARGS
+--test-repo basic.repo
+--test-repo testing:pinning.repo2
+--test-instdb pinning.installed2
+--test-world "c@testing"
+upgrade -a
+@EXPECT
+(1/3) Downgrading b (3 -> 2)
+(2/3) Upgrading a@testing (3 -> 3.1)
+(3/3) Upgrading c@testing (3 -> 3.1)
+OK: 0 MiB in 3 packages
diff --git a/test/pinning9.test b/test/pinning9.test
new file mode 100644
index 0000000..e9836ff
--- /dev/null
+++ b/test/pinning9.test
@@ -0,0 +1,8 @@
+@ARGS
+--test-repo basic.repo
+--test-repo testing:pinning.repo2
+--test-instdb pinning.installed2
+--test-world "c@testing"
+add
+@EXPECT
+OK: 0 MiB in 3 packages