Dbus connection contained in manager client or external #188

jbbjarnason · 2023-10-09T15:29:09Z

Already using this openbmc/sdbusplus#83 with sdbusplus patch, simplifying api

.github/workflows/vcpkg-install-test.yml

libs/confman/inc/public/tfc/confman.hpp

jbbjarnason · 2023-10-09T15:58:46Z

libs/confman/inc/public/tfc/confman/detail/config_dbus_client.hpp

@@ -30,11 +32,13 @@ struct config_property {

 namespace dbus {
 static constexpr std::string_view property_name{ "config" };
-}
+static constexpr std::string_view config_path{ "Config" };


changed path of config, should we change the property_name as well to capital first letter? @SindriTh

Sounds good to me, what does the DBUS naming standard say?

jbbjarnason · 2023-10-09T16:00:09Z

libs/ipc/inc/public/tfc/ipc.hpp

+ auto new_state_filter = [this](value_t&& new_value) {
+ // Here we get unfiltered new value and test whether the value matches the current value
+ auto const& last_value = value();
+ // clang-format off
+ PRAGMA_CLANG_WARNING_PUSH_OFF(-Wfloat-equal)
+ if (!last_value.has_value() || new_value != last_value.value()) {
+ PRAGMA_CLANG_WARNING_POP
+ // clang-format on
+ this->filters_(std::move(new_value));
+ return;
+ }
+ };


maybe this could be moved back, should we?

fyi filters cannot be moved

This does not really make sense to me.
You are comparing a value that is just off the wire with a value
that made it through the filters?

what if invert is applied?

Am I misunderstanding here.

now caching unfiltered value with original new state filter within impl, added separate getter of unfiltered_value and value in ipc

jbbjarnason · 2023-10-09T16:00:59Z

libs/ipc/inc/public/tfc/ipc/details/dbus_slot.hpp

+ void on_set(tfc::stx::invocable<value_t&&> auto&& callback) {
+ interface_->register_method(
+ std::string{ dbus::tags::tinker },
+ [callb = std::forward<decltype(callback)>(callback)](value_t const& set_value) { callb(value_t{ set_value }); });


sdbusplus does not support rvalue callback ...

jbbjarnason · 2023-10-09T16:02:15Z

libs/ipc/src/dbus_client_iface.cpp

- : ipc_manager_client(std::make_shared<sdbusplus::asio::connection>(ctx, tfc::dbus::sd_bus_open_system())) {}
+ : ipc_manager_client(std::make_shared<sdbusplus::asio::connection>(ctx, tfc::dbus::sd_bus_open_system())) {
+ // If the owner of this client does not supply dbus connection we assume we are the sole owner of the connection
+ connection_->request_name(tfc::dbus::make_dbus_process_name().c_str());


if manager client is spawned by context he makes the connection and requests name, not sure what happens if there are more than one who requests the same name? But I guess it won't happen in current code base

But I guess it won't happen in current code base

Famous last words

request_name, indicates that this routine can fail. Maybe we should try it and catch the error?

it throws an error

if (r < 0) { throw exception::SdBusError(-r, "sd_bus_request_name"); }

but I don't see a reason catching it because we cannot continue if it fails right?

libs/confman/src/detail/config_dbus_client.cpp

omarhogni · 2023-10-09T16:32:46Z

libs/confman/src/detail/config_dbus_client.cpp

+ // If the provided interface is created by this class we initialize it here,
+ // otherwise we assume it is taken care of elsewhere
+ if (!interface_name_.empty()) {
+ dbus_interface_->initialize();
+ }


So you create a interface to share with multiple parties.
then one of the parties initalizes the interface?

yes, similar to dbus connection

omarhogni · 2023-10-09T16:35:28Z

libs/ipc/inc/public/tfc/ipc.hpp

@@ -103,25 +110,44 @@ class slot {

 auto operator=(slot const&) -> slot& = delete;

- [[nodiscard]] auto value() const noexcept -> std::optional<value_t> const& { return slot_->value(); }
+ [[nodiscard]] auto value() const noexcept -> std::optional<value_t> const& { return filters_.value(); }

 [[nodiscard]] auto name() const noexcept -> std::string_view { return slot_->name(); }

 [[nodiscard]] auto full_name() const noexcept -> std::string { return slot_->name_w_type(); }


I would look into calling this name_with_type or something more meaningful then full_name.
Maybe unique_name?

I would say name_with_type does not describe the whole truth as it is not strictly name with type, because it go prepended executable&process name

omarhogni · 2023-10-09T16:41:35Z

libs/ipc/src/dbus_client_iface.cpp

- : ipc_manager_client(std::make_shared<sdbusplus::asio::connection>(ctx, tfc::dbus::sd_bus_open_system())) {}
+ : ipc_manager_client(std::make_shared<sdbusplus::asio::connection>(ctx, tfc::dbus::sd_bus_open_system())) {
+ // If the owner of this client does not supply dbus connection we assume we are the sole owner of the connection
+ connection_->request_name(tfc::dbus::make_dbus_process_name().c_str());


request_name, indicates that this routine can fail. Maybe we should try it and catch the error?

omarhogni · 2023-10-09T16:45:11Z

libs/ipc/inc/public/tfc/ipc.hpp

+ auto new_state_filter = [this](value_t&& new_value) {
+ // Here we get unfiltered new value and test whether the value matches the current value
+ auto const& last_value = value();
+ // clang-format off
+ PRAGMA_CLANG_WARNING_PUSH_OFF(-Wfloat-equal)
+ if (!last_value.has_value() || new_value != last_value.value()) {
+ PRAGMA_CLANG_WARNING_POP
+ // clang-format on
+ this->filters_(std::move(new_value));
+ return;
+ }
+ };


This does not really make sense to me.
You are comparing a value that is just off the wire with a value
that made it through the filters?

what if invert is applied?

Am I misunderstanding here.

sonarcloud · 2023-10-10T13:45:49Z

Kudos, SonarCloud Quality Gate passed!

0 Bugs
0 Vulnerabilities
0 Security Hotspots
14 Code Smells

No Coverage information
0.0% Duplication

jbbjarnason force-pushed the one-dbus-connection branch from 2a54b55 to d0a2863 Compare October 9, 2023 15:55

jbbjarnason marked this pull request as ready for review October 9, 2023 15:55

jbbjarnason and others added 20 commits October 9, 2023 15:55

create helper method to generate name for dbus connection

4253f58

reusing same connection within operations exe

ff891ed

small refactor of dbus interface owning for slot

5e5f5e1

broken build but wip

6f36774

exposing injectable dbus interface to config dbus client

ddb751d

exposing injectable dbus interface in confman

ee29e42

wip moving new state detection to filter

8944130

wip moving new state detection to filter

e106cb3

wip going out of scope

8cee350

make it somewhat work

8a07d6b

fix incomplete code

5967676

fix compile errors

3956003

add name to dbus connection

7a562bd

add missing if stmt in cmake

80f6aab

remove files

4c4cde4

Committing clang-format changes

9939946

sonar being stupid

ca05f49

me being stupid ?

f86e4d1

nope sonar being stupid

2103908

format

d0a2863

jbbjarnason commented Oct 9, 2023

View reviewed changes

.github/workflows/vcpkg-install-test.yml Outdated Show resolved Hide resolved

jbbjarnason commented Oct 9, 2023

View reviewed changes

libs/confman/inc/public/tfc/confman.hpp Show resolved Hide resolved

jbbjarnason commented Oct 9, 2023

View reviewed changes

jbbjarnason requested review from omarhogni and SindriTh October 9, 2023 16:02

omarhogni requested changes Oct 9, 2023

View reviewed changes

jbbjarnason added 2 commits October 10, 2023 08:30

Exposing Type as json schema for ipc slots

18f47cd

new state filter works strictly on unfiltered raw value from signal

ef5e4ac

jbbjarnason force-pushed the one-dbus-connection branch from 223c03b to bb30e1c Compare October 10, 2023 09:09

jbbjarnason added 6 commits October 10, 2023 09:09

add getter for unfiltered value

bb30e1c

fix gcc warning

b92f098

disable sonar

3fc29f1

move disable sonar

3e444f0

fix tests

71dce6c

naming

bf5cc59

jbbjarnason requested a review from omarhogni October 10, 2023 09:50

fix operation mode after service name changes

7a67cac

omarhogni approved these changes Oct 10, 2023

View reviewed changes

jbbjarnason added 3 commits October 10, 2023 12:36

fix segfault in test

d69ef4f

fix confman tests

689ef6b

damn boy fix ipc

e2cef61

jbbjarnason force-pushed the one-dbus-connection branch from d0103a0 to f0f76eb Compare October 10, 2023 13:43

Committing clang-format changes

f0f76eb

jbbjarnason merged commit a4f6ca9 into main Oct 10, 2023
16 checks passed

jbbjarnason deleted the one-dbus-connection branch October 10, 2023 14:13

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Dbus connection contained in manager client or external #188

Dbus connection contained in manager client or external #188

jbbjarnason commented Oct 9, 2023 •

edited

Loading

jbbjarnason Oct 9, 2023

SindriTh Oct 9, 2023

jbbjarnason Oct 9, 2023

jbbjarnason Oct 9, 2023

omarhogni Oct 9, 2023

jbbjarnason Oct 10, 2023

jbbjarnason Oct 9, 2023

jbbjarnason Oct 9, 2023

SindriTh Oct 9, 2023

omarhogni Oct 9, 2023

jbbjarnason Oct 10, 2023

omarhogni Oct 9, 2023

jbbjarnason Oct 10, 2023

omarhogni Oct 9, 2023

jbbjarnason Oct 10, 2023

omarhogni Oct 9, 2023

omarhogni Oct 9, 2023

sonarcloud bot commented Oct 10, 2023

Dbus connection contained in manager client or external #188

Dbus connection contained in manager client or external #188

Conversation

jbbjarnason commented Oct 9, 2023 • 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

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

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

sonarcloud bot commented Oct 10, 2023

jbbjarnason commented Oct 9, 2023 •

edited

Loading