Flux request decode raw fix #1144

morrone · 2017-08-08T21:13:40Z

No description provided.

Fix the type of the "data" parameter in the flux_request_decode_raw(). In all uses, "data" is a the address of a pointer, so "void **data" is the correct type. This makes it fairly clear that the buffer will be provided by flux_request_decode_raw() rather than by the caller.

Fix the documentation of the "len" paramter in the flux_request_encode_raw() function. It is an int, not a pointer. Also expand the explanation a little bit.

chu11 · 2017-08-08T21:18:49Z

doc/man3/flux_request_encode.adoc

@@ -16,7 +16,7 @@ SYNOPSIS
                                  const char *json_str);

 flux_msg_t *flux_request_encode_raw (const char *topic,
-                                      void *data, int *len);
+                                      void *data, int len);



I don't think you meant to change flux_request_encode_raw?

oh whoops, nevermind. This is a typo fix, not a change related to the other stuff.

Yeah, just along for the ride, sorry. :)

grondo · 2017-08-08T21:27:17Z

LGTM, nice catch.

chu11 · 2017-08-08T21:31:17Z

I brought up this issue with @garlick awhile back. Looking through old e-mails we never came to a conclusion.

I believe it was initially programmed this way to avoid void ** casting. But I think this is overall more correct/less confusing and worth the void ** casting. But I'll let others chime in.

garlick · 2017-08-08T21:32:27Z

doc/man3/flux_request_decode.adoc

@@ -22,7 +22,7 @@ SYNOPSIS

 int flux_request_decode_raw (const flux_msg_t *msg,
                              const char **topic,
-                              void *data, int *len);
+                              void **data, int *len);


I may be so very wrong about wanting to do this, but it was quite on purpose, and the reason the result parameter is defined as a void * instead of a void ** is so that the user doesn't have to do something like

const char *data; int len; flux_request_decode_raw (msg, NULL, (void **)&data, &len);

in virtually every place this function is used.

I should temper that comment with the observation that now it doesn't seem all that wonderful an idea. What do others think?

For me, when I first looked at:

int flux_request_decode_raw (const flux_msg_t *msg, const char **topic, void *data, int *len);

and compared it to

flux_msg_t *flux_request_encode_raw (const char *topic, const void *data, int len);

My thought was "how come topic & len are in & out parameters, but not data?" You would think it should be a void **.

I think the cast to (void **) is not so bad for increased clarity in the prototype.

If it becomes too onerous a user can create a wrapper function so at most we'd see the case once per file/module.

OK, sounds like @grondo already acked so let's fix this. Want to go ahead and also fix:

int flux_msg_get_payload (const flux_msg_t *msg, int *flags, void *buf, int *size); int flux_mrpc_get_raw (flux_mrpc_t *mrpc, void *data, int *len); int flux_rpc_get_raw (flux_future_t *f, void *data, int *len); int flux_response_decode_raw (const flux_msg_t *msg, const char **topic, void *data, int *len);

for consistency?

👍 forgot about those.

Sure, I can work on those others.

@chu11: just changed prototypes for treeobj_decode_val() and flux_kvs_lookup_get_raw() in the branch we're working on and pushed.

grondo · 2017-08-08T21:49:21Z

Some checks in Travis failed and I'm restarting them. I saw another case of "write error" during the distcheck packaging phase, and a hang due to failure in the faketime tests. All known issues obviously not related to this PR.

As discussed in flux-framework#1144, function result parameters like the one in treeobj_decode_val() should probably be void ** not void *.

As discussed in flux-framework#1144, function result parameters like the one in flux_kvs_lookup_get_raw() should probably be void ** not void *.

coveralls · 2017-08-08T22:09:39Z

Coverage decreased (-0.02%) to 77.963% when pulling 20b505f on morrone:flux_request_decode_raw_fix into 0abcd67 on flux-framework:master.

The function want to return a pointer to a buffer in one of its parameters, so that parameter should be a void ** rather than a void *.

codecov-io · 2017-08-08T22:55:45Z

Codecov Report

Merging #1144 into master will increase coverage by 0.02%.
The diff coverage is 92.85%.

@@            Coverage Diff             @@
##           master    #1144      +/-   ##
==========================================
+ Coverage   77.61%   77.63%   +0.02%     
==========================================
  Files         157      157              
  Lines       28679    28680       +1     
==========================================
+ Hits        22260    22267       +7     
+ Misses       6419     6413       -6

Impacted Files	Coverage Δ
src/common/libflux/mrpc.c	`86.32% <0%> (ø)`	⬆️
src/modules/kvs/kvs.c	`63.45% <100%> (+0.78%)`	⬆️
src/cmd/builtin/content.c	`73.87% <100%> (ø)`	⬆️
src/common/libflux/content.c	`86.66% <100%> (ø)`	⬆️
src/broker/content-cache.c	`73.27% <100%> (-1.3%)`	⬇️
src/common/libflux/request.c	`89.74% <100%> (+1.28%)`	⬆️
src/broker/log.c	`69.64% <100%> (ø)`	⬆️
src/common/libflux/response.c	`83.73% <100%> (ø)`	⬆️
src/common/libflux/rpc.c	`94.21% <100%> (+0.82%)`	⬆️
src/common/libflux/message.c	`81.76% <100%> (+0.58%)`	⬆️
... and 16 more

morrone · 2017-08-08T22:57:42Z

OK, I added a couple of commits to this pull requires to address these five functions:

flux_msg_get_payload
flux_rpc_get_raw
flux_response_decode_raw
flux_mrpc_get_raw
flux_content_load_get

coveralls · 2017-08-08T23:26:26Z

Coverage decreased (-0.04%) to 77.945% when pulling 3e8cb3b on morrone:flux_request_decode_raw_fix into 0abcd67 on flux-framework:master.

chu11 · 2017-08-08T23:29:19Z

doc/man3/tmrpc_then

@@ -0,0 +1,228 @@
+#! /bin/bash


Did you accidentally git add this file?

Doh. Fixed.

The functions want to return a pointer to a buffer in one of its parameters, so that parameter should be a void ** rather than a void *. The following functions are converted: flux_rpc_get_raw flux_response_decode_raw flux_content_load_get flux_mrpc_get_raw

coveralls · 2017-08-09T01:47:58Z

Coverage decreased (-0.05%) to 77.931% when pulling 45f5341 on morrone:flux_request_decode_raw_fix into 0abcd67 on flux-framework:master.

grondo · 2017-08-10T15:15:58Z

restarted travis since it failed to update status of the PR and GitHub wouldn't let me press the button.

As discussed in flux-framework#1144, function result parameters like the one in treeobj_decode_val() should probably be void ** not void *.

As discussed in flux-framework#1144, function result parameters like the one in flux_kvs_lookup_get_raw() should probably be void ** not void *.

In flux_future_get, convert void * parameter to void ** parameter, consistent to fixes done in flux-framework#1144 and flux-framework#1212. Fixes flux-framework#1602

morrone added 2 commits August 8, 2017 14:05

man3: Fix prototype of flux_request_encode_raw()

20b505f

Fix the documentation of the "len" paramter in the flux_request_encode_raw() function. It is an int, not a pointer. Also expand the explanation a little bit.

morrone requested review from chu11 and garlick August 8, 2017 21:14

chu11 reviewed Aug 8, 2017

View reviewed changes

garlick reviewed Aug 8, 2017

View reviewed changes

grondo mentioned this pull request Aug 8, 2017

flux_open: No such file or directory #677

Closed

garlick added a commit to garlick/flux-core that referenced this pull request Aug 8, 2017

libkvs/treeobj: treeobj_decode_val use void **

cdc39e2

As discussed in flux-framework#1144, function result parameters like the one in treeobj_decode_val() should probably be void ** not void *.

garlick added a commit to garlick/flux-core that referenced this pull request Aug 8, 2017

libkvs/lookup: lookup_get_raw use use void **

73dc493

As discussed in flux-framework#1144, function result parameters like the one in flux_kvs_lookup_get_raw() should probably be void ** not void *.

libflux/message: Change void * to void ** in flux_msg_get_payload

c360963

The function want to return a pointer to a buffer in one of its parameters, so that parameter should be a void ** rather than a void *.

chu11 reviewed Aug 8, 2017

View reviewed changes

grondo merged commit 1665912 into flux-framework:master Aug 10, 2017

morrone deleted the flux_request_decode_raw_fix branch August 10, 2017 22:15

chu11 pushed a commit to chu11/flux-core that referenced this pull request Aug 15, 2017

libkvs/treeobj: treeobj_decode_val use void **

47e8126

As discussed in flux-framework#1144, function result parameters like the one in treeobj_decode_val() should probably be void ** not void *.

chu11 pushed a commit to chu11/flux-core that referenced this pull request Aug 15, 2017

libkvs/lookup: lookup_get_raw use use void **

6f8b402

As discussed in flux-framework#1144, function result parameters like the one in flux_kvs_lookup_get_raw() should probably be void ** not void *.

grondo mentioned this pull request Aug 23, 2017

0.8.0 Release #1160

Closed

chu11 mentioned this pull request Jul 24, 2018

flux_future_get() should take const void ** #1602

Closed

chu11 mentioned this pull request Jul 27, 2018

flux_future_get: void * -> const void ** #1611

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Flux request decode raw fix #1144

Flux request decode raw fix #1144

morrone commented Aug 8, 2017

chu11 Aug 8, 2017

chu11 Aug 8, 2017

morrone Aug 8, 2017

grondo commented Aug 8, 2017

chu11 commented Aug 8, 2017

garlick Aug 8, 2017

garlick Aug 8, 2017

chu11 Aug 8, 2017 •

edited

Loading

grondo Aug 8, 2017

garlick Aug 8, 2017

grondo Aug 8, 2017

morrone Aug 8, 2017

garlick Aug 8, 2017

grondo commented Aug 8, 2017

coveralls commented Aug 8, 2017

codecov-io commented Aug 8, 2017 •

edited

Loading

morrone commented Aug 8, 2017

coveralls commented Aug 8, 2017

chu11 Aug 8, 2017

morrone Aug 9, 2017

coveralls commented Aug 9, 2017 •

edited

Loading

grondo commented Aug 10, 2017

Flux request decode raw fix #1144

Flux request decode raw fix #1144

Conversation

morrone commented Aug 8, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

grondo commented Aug 8, 2017

chu11 commented Aug 8, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chu11 Aug 8, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

grondo commented Aug 8, 2017

coveralls commented Aug 8, 2017

codecov-io commented Aug 8, 2017 • edited Loading

Codecov Report

morrone commented Aug 8, 2017

coveralls commented Aug 8, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

coveralls commented Aug 9, 2017 • edited Loading

grondo commented Aug 10, 2017

chu11 Aug 8, 2017 •

edited

Loading

codecov-io commented Aug 8, 2017 •

edited

Loading

coveralls commented Aug 9, 2017 •

edited

Loading