On Sun, 03 May 2020 01:09:15 +0200 nicoo <[email protected]> wrote: > Package: src:cmake > Version: 3.16.3-2 > Severity: important > Tags: upstream > Control: clone -1 -2 > Control: reassign -2 debhelper > Control: retitle -2 debhelper: Do not set CMAKE_AUTOGEN_VERBOSE=NO > Control: severity -2 normal > Control: found -2 12.10 > Control: found -2 13 > Control: affects -1 debhelper > > Dear maintainers, > > > CMake compiles /usr/share/cmake-*/Modules/CMakeCXXCompilerABI.cpp during the > build of a C++ projects, several times, and ignoring the CXXFLAGS. > This behaviour is likely defined in > > /usr/share/cmake-*/Modules/{CMakeTestCXXCompiler,Compiler/{Intel,GNU,QCC}}.cmake > > In addition to being a (presumably benign?) correctness issue, this is about > to > become a notable inconvenience for maintainers of C++ packages in Debian whose > buildsystem is CMake: > - starting in debhelper 12.10 (uploaded 2020-03-24), CMAKE_AUTOGEN_VERBOSE=ON > is set by default for all packages that use CMake ; > - CMake then logs the compiler invocations for CMakeCXXCompilerABI.cpp; > - blhc (build log hardening checker) raises (correctly so) an error, > as C++ code was built without the required flags. > - the blhc failure is surfaced to the maintainer in many places: > Salsa CI, tracker.d.o, qa.d.o, ... > > After discussion with nthykier, he suggested rolling back the change in dh > (commit 2f88a603001a32b596389b8b031489960e30a9e9) until this issue is fixed > in CMake. As such I am leaving a copy of this bug report against dh. > > > Many thanks to Andreas Noteng, whose package scantailor-advanced (which I > am sponsoring) is where I first encountered the issue, and who bissected > the problem down to that specific commit in debhelper. > > > Best, > > nicoo > > [...]
Hi Коля Гурьев, Just a heads up to make sure you are aware of bug as you made the MR to add the apply that flag (https://salsa.debian.org/debian/debhelper/-/commit/2f88a603001a32b596389b8b031489960e30a9e9). My initial take on this is to roll back the flag until cmake supports it better. @Nicolas: Do you know if bhlc checks are commonly done in -backports as well? I would like to know whether a cmake fix for this in unstable is sufficient or we should wait until cmake is fixed in stable(-backports). ~Niels
