Merge templated RPC migration into develop #176
Conversation
* Add cross-compile (un)install targets
* change `NULL` to `nullptr` * remove some seemingly unnecessary `#include`s
* Added placeholders for several needed serialize methods * serializers for 8 and 16 bit unsigned types * serializers for float using simple `reinterpret_cast` **may not work** * Remove `GETLOCALARGS` macro * Cleanup `#include` directives * Cleanup doxy comment styles * Clean style for braces and `struct` names
* Doesn't compile because not all submodules have been migrated
* doesn't compile because not all `amc` submodules have been migrated
* doesn't compile because not all `amc` submodules have been migrated
* doesn't compile because not all `amc` submodules have been migrated
* update structs for enuns to strongly typed enum struct
* remove vestigial stdsprintf * cleanup commented includes
* add spec.template and put packageinfo in correct location * correction of types, serializer for `size_t`
* Update config submodule and use `https` instead of `ssh` * Add gitlab-ci config * Update gitignore
[feature] templated RPC + package structure
utils::regExists closes the LMDB when it exits. Whenever the db_res was used, it was referring to a dangling pointer to a (non-existent) memory-mapped address. utils::regExists has been modified to return the content of the LMDB key as a std::string if it exists, avoiding the previsouly described issue. During the adaptation of utils::writeRawAddress, utils::readRawAddress, utils utils::writeRawReg, utils::readRawReg, utils::writeReg and utils::readReg, the following internal functions have been removed: * utils::writeAddress * utils::readAddress Basic const-ification has been applied on the way.
* Update the config submodule to the latest commit in feature/common-ci * Disable the x86_64 build while the GLIB sysroot is not ready
int is not a (de)serializable type by default. Since the signed behavior is not needed, just remove it with the unsigned version.
Replacing it with a std::vector<std::string>
…ted-rpc Minimal working version of the templated RPC methods
…d readRemoteBlock
|
Historically, for all merges done via PR/MR (except for those in throwaway branches, e.g., what was supposed to be the "short-term" legacy code) should have a merge commit to mark the PR/MR through which the merge happened (history and automatic changelog are two of many reasons to do this). |
It makes a lot sense for regular PR/MR. Does it also make so much sense for central branches merges? Particularly in this case where the development was stalled in However, that rule has been very loosely applied. If I'm not mistaken, non fast-forward merges are even forbidden in some repositories, or, at least, not the default option.
👍 👍 👍 |
- adding [[deprecated]] attributes to single register access exports - removing export of readRemoteBlock as it is indeed not used by cmsgemos - getting order in include/server/utils.h
Export read and write functions as RPC methods as a temporary compatibility layer
|
Merging as is will allow to loosely keep the commit history after deletion of the |
Description
Following what was done in
xHALand as described in #172, this PR aims at makingdevelopthe main development branch.The only active development is around the templated RPC, but the target branch is different between different repositories (
developinxhalandcmsgemos, butfeature/templated-rpcin this repository). It is somehow confusing.The
developbranch ofcmsgemosis going to start using the templated RPC framework, so it makes sense to include the templated RPC functionalities intodevelop.Moreover, some of the latest clean up PR would be target
developrather thanfeature/templated-rpc-methods, but are (going to be) merged intofeature/templated-rpc-methodsto avoid any divergences.Since fast forward merge is better done from the command line (committers are preserved), I'm going to push the change this evening if there is no strong opposition.
Types of changes
Motivation and Context
The templated RPC are the present of the CTP7 modules.