boost: Adds with_stacktrace_noop option #7503
Conversation
|
|
|
Sorry, the build is only launched for Access Request users. You can request access writing in this issue. |
|
I detected other pull requests that are modifying boost/all recipe:
This message is automatically generated by https://github.com/ericLemanissier/conan-center-conflicting-prs so don't hesitate to report issues/improvements there. |
| @@ -1567,7 +1569,8 @@ def filter_transform_module_libraries(names): | |||
| self.cpp_info.components["stacktrace_backtrace"].defines.append("BOOST_STACKTRACE_USE_BACKTRACE") | |||
| self.cpp_info.components["stacktrace_backtrace"].requires.append("libbacktrace::libbacktrace") | |||
|
|
|||
| self.cpp_info.components["stacktrace_noop"].defines.append("BOOST_STACKTRACE_USE_NOOP") | |||
There was a problem hiding this comment.
stacktrace_noop is a header-only library.
So the component is always available (when stacktrace is enabled)
Sure, you can add an option to specifically disable it, but is it worth it?
If you push further, please also do the following:
- in
configure, dodel self.options.with_stcktrace_noopif stacktrace is disabled. - Also
dodel self.info.options.with_stacktrace_noopinpackage_id` (this can be done unconditionally)
There was a problem hiding this comment.
What do you mean stacktrace_noop is a header-only library? From what I understand, the noop is an implementation detail that is activated with the BOOST_STACKTRACE_USE_NOOP define. I'm new to the conan package, so is there another way to disable the noop define? Thanks!
There was a problem hiding this comment.
Look in the main recipe, there is only one noop match:
conan-center-index/recipes/boost/all/conanfile.py
Line 1570 in 4f4a376
This component is set when stacktrace is enabled:
conan-center-index/recipes/boost/all/conanfile.py
Line 1552 in 4f4a376
It does not add any libs.
self.cpp_info.components["stacktrace_noop"].libs = [...] is not present.
When writing the stacktrace tests, I found out that noop is the fallback backend of stacktrace.
If no implementation can be found, then nothing (=no-op) will be done.
The only way to disable no-op, is to make sure that another backend is available and used.
There was a problem hiding this comment.
I see, then shouldn't the backtrace backend (which is True by default) disable no-op?
There was a problem hiding this comment.
No? no-op is always available, even when the addr2line/libbacktrace backend is built.
All three boost_stacktrace backends (noop/addr2line/libbacktrace) can be available in the same pacakge, just not used in the same target when using the cmake targets.
You can switch between the backends by carefully setting the definitions and linking to the correct libraries.
There was a problem hiding this comment.
Will do, thanks!
Also, you mentioned that cmake would be deprecated as a generator. Is that for conan as a whole or just boost?
There was a problem hiding this comment.
I mean only for Conan.
conanbuiodinfo.cmake will be a thing of the past.
I'm taking about Conan 2.0, coming in the (near) future.
There was a problem hiding this comment.
I see, thanks!
As an aside, what build system would you recommend outside of cmake these days?
There was a problem hiding this comment.
To be honest, I use cmake for all of my private and professional c/c++ projects.
All other build systems have disadvantages that are showstoppers for me.
Though, a simple Makefile might be useful for something that has nothing to do with compiling sources.
[Meson}(https://mesonbuild.com/) is used by some OSS projects.
But I find its documentation hard to read + it is opiniated.
It has a potential though.
A meson syntax with a cmake backend would be very nice..
The grandfather of build systems for OSS is of course autotools (autoconf + automake + libtool).
But that is geared mostly towards Linux + gcc. It can work with MSVC on Windows, but it's harder.
If you really don't care about portability, you can also of course just use Visual Studio projects (.snl and .vcxproj files).
So in general, I can't really recommend anything but cmake.
There was a problem hiding this comment.
Thank you, appreciate your insights!
| add_executable(stacktrace_noop_exe stacktrace.cpp) | ||
| if(WITH_STACKTRACE_NOOP) | ||
| target_compile_definitions(stacktrace_noop_exe PRIVATE TEST_STACKTRACE_IMPL=4) | ||
| else() | ||
| # BOOST_STACKTRACE_USE_NOOP is not defined in this case, so we set it here only for this target | ||
| target_compile_definitions(stacktrace_noop_exe PRIVATE TEST_STACKTRACE_IMPL=4 PRIVATE BOOST_STACKTRACE_USE_NOOP) | ||
| endif() |
There was a problem hiding this comment.
Please fix your formatting (space, 4chars).
Also, you always test noop. This is because noop is header-only.
So this shows why I think this noop option is non-sensical.
The only thing it does is enable/disable exposure of Boost::stacktrace_noop.
| else() | ||
| # BOOST_STACKTRACE_USE_NOOP is not defined in this case, so we set it here only for this target | ||
| target_compile_definitions(stacktrace_noop_exe PRIVATE TEST_STACKTRACE_IMPL=4 PRIVATE BOOST_STACKTRACE_USE_NOOP) | ||
| endif() | ||
| target_link_libraries(stacktrace_noop_exe PRIVATE Boost::stacktrace_noop) |
There was a problem hiding this comment.
This target will not be available when stacktrace_noop is disabled.
ghost
closed this
Specify library name and version: boost/all
This is to add a with_stacktrace_noop option to boost::stacktrace. Currently, the BOOST_STACKTRACE_USE_NOOP define is set always and there isn't a way to disable it without modifying downstream generated build configs.
conan-center hook activated.