rpk: segfault on errors in topic add-partitions

[jcsp]

Testing with tip of dev on a system with lots of partitions.

$ rpk topic add-partitions scale_10k -n20000
TOPIC  ERROR
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x18 pc=0xc4aecf]

goroutine 1 [running]:
github.com/redpanda-data/redpanda/src/go/rpk/pkg/cli/cmd/topic.NewAddPartitionsCommand.func1(0xc000493680, {0xc0004c0680, 0x1, 0x2})
	/home/vectorized/redpanda/src/go/rpk/pkg/cli/cmd/topic/add_partitions.go:65 +0x48f
github.com/spf13/cobra.(*Command).execute(0xc000493680, {0xc0004c0660, 0x2, 0x2})
	/home/vectorized/redpanda/vbuild/go/1.17.8/pkg/mod/github.com/spf13/cobra@v1.1.3/command.go:856 +0x5f8
github.com/spf13/cobra.(*Command).ExecuteC(0xc000458f00)
	/home/vectorized/redpanda/vbuild/go/1.17.8/pkg/mod/github.com/spf13/cobra@v1.1.3/command.go:960 +0x3ad
github.com/spf13/cobra.(*Command).Execute(...)
	/home/vectorized/redpanda/vbuild/go/1.17.8/pkg/mod/github.com/spf13/cobra@v1.1.3/command.go:897
github.com/redpanda-data/redpanda/src/go/rpk/pkg/cli/cmd.Execute()
	/home/vectorized/redpanda/src/go/rpk/pkg/cli/cmd/root.go:117 +0x630
main.main()
	/home/vectorized/redpanda/src/go/rpk/cmd/rpk/main.go:15 +0x17

Looks like this is a null pointer dereference on err, where it means to read e.

[jcsp]

Would be interested to see what a regression test looks like for this -- I can imagine a short ducktape test that e.g. tries to add an unfeasible number of partitions to generate an error, but maybe there's enough unit testing infrastructure to do it that way.

[twmb]

This is a simple bug that uses the wrong variable internally, where the variable is nil but the one it's meant to be using is not