Re: [boost] [system] Would it be possible to trial a breaking change to Boost.System and see what happens?

Subject: Re: [boost] [system] Would it be possible to trial a breaking change to Boost.System and see what happens?
From: Andrey Semashev (andrey.semashev_at_[hidden])
Date: 2018-01-15 10:37:45

On 01/15/18 03:01, Niall Douglas via Boost wrote:
> 
> Is anybody actually generating a totally unique message per invocation
> for the same error code which couldn't be cached and reused? I've never
> seen it yet. Almost all use of message() is for printing to human
> readable output.

libc++ uses strerror_r or snprintf to generate error messages (see 
system_error.cpp[1]). libstdc++ uses strerror[2] (which seems like a bug 
as it is not thread-safe). strerror_r implementation in glibc may 
dynamically generate the error message, although most of the time it 
returns one of many constant strings. Either way, users of strerror_r 
have to assume the result is dynamically generated.

You didn't mention how you would solve the problem of thread 
synchronization to the static storage or cache. In case of a hash map, 
how would you be removing entries from it? Because if you don't, you're 
wasting memory.

[1]: https://llvm.org/svn/llvm-project/libcxx/trunk/src/system_error.cpp
[2]: 
https://github.com/gcc-mirror/gcc/blob/master/libstdc%2B%2B-v3/src/c%2B%2B11/system_error.cc