added convience functions#15
Conversation
Expose fd in Socket to allow non-blocking operations
This adds some missing functions for sockets and pulls the TriggerFd class in from Subspace to make it usable in other projects
Fixed some pedantic compiler warnings
Fixed gcc c++17 error about field order in aggregate init
Add coroutine-aware Read/Write functions
Virtual sockets. I'm not sure this is fully working yet, but I need it for subspace
…t_names Safer handling of unix socket address names
|
realized logger has capital function names. Do you want me to change to |
|
Yes please, Google style FTW |
| Log(LogLevel::kERROR, std::move(str)); | ||
| } | ||
| inline void Fatal(std::string&& str) { | ||
| Log(LogLevel::kFATAL, std::move(str)); |
There was a problem hiding this comment.
There isn't really a benefit to these rvalue overloads when the underlying function does not swallow up that object and thus can only really use it as a const-ref.
| if (!_enabled || LogLevel::kVERBOSE_DEBUG < min_level_) { | ||
| return; | ||
| } | ||
| Log(LogLevel::kVERBOSE_DEBUG, fmt::format(fmt, std::forward<Ts>(args)...)); |
There was a problem hiding this comment.
If this file is going to be using fmt, it needs to include the relevant headers and dependency in the bazel target.
| if (!_enabled || LogLevel::kDBG < min_level_) { | ||
| return; | ||
| } | ||
| Log(LogLevel::DBG, fmt::format(fmt, std::forward<Ts>(args)...)); |
There was a problem hiding this comment.
It's a weird that when the user calls Log(lvl, str, a, b, c) (i.e., the vararg version) it gets formatted, under-the-hood, with the "printf" formatting, but then, if they call Info(str, a, b, c) it gets formatted with the python-style fmt specifications.
N.B.: I much prefer the fmt version, but I think it needs to be consistent.
There was a problem hiding this comment.
I like the idea of this. In fact, in my own company's logging library, we have similar perfect forwarding of the fmt+args down to the logging functions to present the user with a similarly convenient syntax.
One big design feature that I would very much recommend (and might be able to provide the code for), is to perfectly forward all the way down to the implementation of Log / VLog functions because that will allow for logging functions that do not require any dynamic memory allocation, which is very handy in real-time and signal handler contexts, and quite easy to do with the fmt library. I did this very easily, and I can remove malloc/free in some contexts and still log messages to stdout/stderr.
But, this will require rewriting Log/VLog completely. So, maybe something for @dallison to think about.
There was a problem hiding this comment.
thats a good call. Will update. Thanks
4 participants
Added convience function which avoid str format until after log has been checked.
goes to