modules/kvs: Log to flux not stderr

[chu11]

Replace lingering log calls that log to stderr with calls that
log to flux logging facility. Requires passing optional flux_t
handle to internal lookup and commit APIs.

Remove lingering unnecessary includes of log.h throughout.

Fixes #1165

[codecov-io]

Codecov Report

Merging #1190 into master will increase coverage by 0.01%.
The diff coverage is 60%.

@@            Coverage Diff             @@
##           master    #1190      +/-   ##
==========================================
+ Coverage   77.82%   77.83%   +0.01%     
==========================================
  Files         158      158              
  Lines       29288    29295       +7     
==========================================
+ Hits        22794    22803       +9     
+ Misses       6494     6492       -2

Impacted Files	Coverage Δ
src/modules/kvs/kvs_util.c	87.75% <ø> (ø)	⬆️
src/modules/kvs/kvs.c	63.43% <100%> (+0.26%)	⬆️
src/modules/kvs/lookup.c	84.19% <55.55%> (-0.5%)	⬇️
src/modules/kvs/commit.c	77.41% <60%> (+0.06%)	⬆️
src/common/libkvs/kvs_commit.c	86.66% <0%> (-6.67%)	⬇️
src/common/libflux/content.c	86.66% <0%> (-3.34%)	⬇️
src/modules/connector-local/local.c	74.27% <0%> (-1.28%)	⬇️
src/common/libflux/mrpc.c	85.15% <0%> (-1.18%)	⬇️
src/broker/modservice.c	79.61% <0%> (-0.98%)	⬇️
src/common/libflux/rpc.c	93.38% <0%> (-0.83%)	⬇️
... and 17 more

[coveralls]

Coverage decreased (-0.01%) to 78.147% when pulling 0555fc8 on chu11:issue1165 into 8443eff on flux-framework:master.

[chu11]

codecov diff is unsurprisingly bad, as all changes are "impossible" to hit error cases

[garlick]

Yeah, I think we could ignore the coverage here.

I wonder if it would be a good idea to submit a mini-PR for the logging code before this one containing a change like this, so we can avoid the test for (h != NULL) at each log call?

diff --git a/src/common/libflux/flog.c b/src/common/libflux/flog.c
index 8eae5cb..9944464 100644
--- a/src/common/libflux/flog.c
+++ b/src/common/libflux/flog.c
@@ -134,7 +134,7 @@ done:
 
 int flux_vlog (flux_t *h, int level, const char *fmt, va_list ap)
 {
-    logctx_t *ctx = getctx (h);
+    logctx_t *ctx;
     int saved_errno = errno;
     uint32_t rank;
     int len;
@@ -143,7 +143,11 @@ int flux_vlog (flux_t *h, int level, const char *fmt, va_list ap)
     struct stdlog_header hdr;
     int rpc_flags = FLUX_RPC_NORESPONSE;
 
-    if (!ctx) {
+    if (!h) {
+        return vfprintf (stderr, fmt, ap);
+    }
+
+    if (!(ctx = getctx (h))) {
         errno = ENOMEM;
         goto fatal;
     }

I guess probably there'd be a bit more to it - man page update, a test, maybe include level in the output, interpreted.

[chu11]

@garlick I hadn't thought about that, but I think it's a good idea. I'll take a look at that first.

[chu11]

re-pushed w/ updates given #1192

[garlick]

might want to update this comment

[chu11]

oops! thanks

[coveralls]

Coverage increased (+0.007%) to 78.184% when pulling 2be2edd on chu11:issue1165 into 3cd405f on flux-framework:master.

[garlick]

Thanks @chu11, this looks good.