glaze: Warnings & Errors in various configurations
Since I’m maintaining the build2 package of glaze I just ran it through the (free) CI servers and figured I’ll let you know which has warnings & which has errors (in case you’d rather just tackle them in one go).
See build status here for v2.6.8.
Edit.
Obviously you can ignore the warnings coming from Eigen headers.
I should also mention that there appears to be a sporadic error in the jsonrpc_test on Windows (unless it’s something wrong specifically with my build2 package).
About this issue
- Original URL
- State: closed
- Created 2 years ago
- Comments: 16 (16 by maintainers)
I’m moving to a new version of ut to avoid issues in the library, which should solve the problem seen in the jsonrpc tests. In #1041
I just merged in #1039, which removes most the warnings across the compilers.
To explain a bit more why avoiding unreachable is a challenge: If Glaze were to use exceptions we could easily make local lambda functions that wrap up behavior and then just call these functions in the appropriate
if constexprbranches. However, because Glaze does not use exceptions internally any lambda function breaks our ability to return from the function at hand, we just return from the lambda. This requires an additional error check after the lambda is called, to determine if an error occurred in the lambda, instead of just exiting from the function. So, we lose performance by using lambdas.This is a problem that requires deterministic and non-allocating exception support in C++. There have been proposals to improve exception handling in C++, but I think the broader community needs to see this as more of a critical issue.
Yeah, I saw those. Working on them now.
I’ll take another look at whether I should refactor the code or just locally disable the warning. Thanks for pointing these out.
Clang 15 should build. I run actions with 16, 17, & 18
So, I’ve tried multiple times to fix the MSVC warning for
<expected>, but older versions of Clang break with these kinds of checks. If you’re able to make a pull request that passes, that would be great, but otherwise we’ve just been living with these warnings and they’ll go away in C++23 and beyond.Yeah,
boost-uthas some major issues. I’ve been trying to get them to fix some random segfaults from simply including the header. I’ve developed a simpler ut library that matches the core API of boost-ut that I will probably switch over to.