Kangping

或者我晚点fix～

确实是应该写一个定制的 _**string\_t\_to\_int**_，我太懒了。 状态机的话，严谨的测试需要很多case，这个工作量还是挺多的。 如果对这些感兴趣的话，可以看看nginx的代码。我也是从nginx学过来的，我省去了HTTP的许多逻辑，但是整体的架构、优化的手段，基本都能够在nginx里面找到。推荐你有时间也可以看看，对编程修养的提高还是挺有帮助的（nginx也不过20000 - 30000 loc）～

> > @abtink What about adding unit tests to verify this change covers the described corner cases? > > Good idea @wgtdkp. I added a new test case in `test_routing_manager`...

> This commit updates `NetworkData::Publisher` to reserve a set of prefix entries for use by `RoutingManager` only. This allow us to simplify how `RoutingManager` publishes route prefixes in Network Data...

> > This doesn't guarantees that RoutingManager call always successfully publish the route for the on-link prefixes, because the network data registration can be rejected by the leader for various...

> There seems no real use case for `multiple OT instances` ot-esp32 project needs to dynamically allocate the OT instance.

@lancersky Can you clarify if below is true for your SRP client device: the device has been factory reset but the same value is used to construct the SRP host...

@lmnotran Thanks for this change! The pretty check is failing, would you help resolve this? (You will need to format the c++ code and upload the new changes)

> The macOS build is failing because we was not supporting Border Routing on macOS platforms. Will address this later. Updated the CL to skip SRP Server and Border Routing...

``` sh: ipset: command not found ``` It complains that the `ipset` command is not found on your device. try to install it before starting otbr-agent