Fixes broken migration to klog

[gman0]

Fixes #187

Bug introduced in #150 (quite insufficient testing - plugins would panic immediately, #168) by not playing nicely with vendor'd packages that still use glog (github.com/kubernetes-csi/drivers/pkg/csi-common).

This then continued on in #169 which fixed the panic, but skipping klog.InitFlags() means logtostderr cmd flag doesn't get propagated to klog.

This PR reintroduces klog.InitFlags() in a hopefully less conflicting manner with glog.

[gman0]

cc @rootfs @humblec

[rootfs]

make a util function for this?

[gman0]

will do

[gman0]

I'm getting

E0212 15:15:23.815436    5850 log.go:38] failed to set log_backtrace_at flag: syntax error: expect file.go:234

if I put it in the ceph-csi/pkg/util package - I'm not 100% sure why...i guess because it's in another package?

[gman0]

Nevermind, should be working now...I had to ignore error checking when copying args from glog to klog though.

[rootfs]

thanks @gman0

[humblec]

@gman0 Thanks.. One question here is, the klogInitFlags() has to happen before flags.set https://github.com/ceph/ceph-csi/pull/188/files#diff-48c9752907803b1df4780beae292b547R38, otherwise it will be NO OP , Isnt it ?

[gman0]

@humblec I don't think it makes a difference in this case, because glog already initializes the cmd flags:

First, glog initializes the global command-line FlagSet in its init() function:

ceph-csi/vendor/github.com/golang/glog/glog.go

Lines 398 to 404 in 1748629

	func init() {
	flag.BoolVar(&logging.toStderr, "logtostderr", false, "log to standard error instead of files")
	flag.BoolVar(&logging.alsoToStderr, "alsologtostderr", false, "log to standard error as well as files")
	flag.Var(&logging.verbosity, "v", "log level for V logs")
	flag.Var(&logging.stderrThreshold, "stderrthreshold", "logs at or above this threshold go to stderr")
	flag.Var(&logging.vmodule, "vmodule", "comma-separated list of pattern=N settings for file-filtered logging")
	flag.Var(&logging.traceLocation, "log_backtrace_at", "when logging hits line file:N, emit a stack trace")

...which means the flag.Set() in this PR actually sets the logtostderr flag for glog, not klog => it's not a NOOP.

After the values are parsed by flag.Parse(), a new FlagSet (a separate one, for klog only) is created and klog stores its cmd args here (that's why it doesn't panic anymore, it uses a separate flagset - no duplicate flags):

ceph-csi/vendor/k8s.io/klog/klog.go

Lines 407 to 419 in 1748629

	func InitFlags(flagset *flag.FlagSet) {
	if flagset == nil {
	flagset = flag.CommandLine
	}
	flagset.StringVar(&logging.logDir, "log_dir", "", "If non-empty, write log files in this directory")
	flagset.StringVar(&logging.logFile, "log_file", "", "If non-empty, use this log file")
	flagset.BoolVar(&logging.toStderr, "logtostderr", false, "log to standard error instead of files")
	flagset.BoolVar(&logging.alsoToStderr, "alsologtostderr", false, "log to standard error as well as files")
	flagset.Var(&logging.verbosity, "v", "log level for V logs")
	flagset.BoolVar(&logging.skipHeaders, "skip_headers", false, "If true, avoid header prefixes in the log messages")
	flagset.Var(&logging.stderrThreshold, "stderrthreshold", "logs at or above this threshold go to stderr")
	flagset.Var(&logging.vmodule, "vmodule", "comma-separated list of pattern=N settings for file-filtered logging")
	flagset.Var(&logging.traceLocation, "log_backtrace_at", "when logging hits line file:N, emit a stack trace")

Lastly, the values from glog flagset are copied over to the values of klog flagset. Granted, this is hardly ideal because there are some flags that klog has and glog hasn't, and those cannot be set (because glog manages the global cmd flagset) but AFAIK there's no way around this, unless csi-common moves to klog or we move away from csi-common.

[humblec]

@gman0 The csi-common migration to klog is under review at the same repo by PR #144 , may be getting it done first would be the right choice. I have to do some changes in that PR and I am on it.

[gman0]

PTAL @rootfs @humblec

CI is failing on an unrelated file:

$ scripts/lint-text.sh --require-all
mdl --style scripts/mdl-style.rb ./README.md 
./README.md:27: MD012 Multiple consecutive blank lines

may be fixed in another PR

[rootfs]

rerun CI and it passed. megify failed to merge, will figure out that later. thanks @gman0