Makefile.am, configure.ac: fix hard coded static and include path #1026
Conversation
vjardin
force-pushed
the
fix_br_makefile
branch
from
6562926
to
30c4bb2
Compare
|
Side note : without this patch, it cannot link because |
I do not understand why some -static are enforced while dynamic linkage could be used. We should not have any -static in order to be able to link for the targets.
It is not compliant with cross compilation, it will lead to the following: aarch64-buildroot-linux-gnu-g++ -c -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -O2 -g0 -D_FORTIFY_SOURCE=1 -DHAVE_TERMIOS_H -DHAVE_SYS_PCI_H -isystem /usr/include/libxml2 -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 conftest.cpp aarch64-buildroot-linux-gnu-g++: ERROR: unsafe header/library path used in cross-compilation: '-isystem' '/usr/include/libxml2' configure.ac: proper probe libxml2 libxml2: CFLAGS and CXXFLAGS are used Makefiles are not using xml_CFLAGS
the include path cannot be hardcoded with: /usr/include/curl/
engine is exported by openssl/engine.h ; if not, we can have some compilation warnings of some undefined symbols.
vjardin
force-pushed
the
fix_br_makefile
branch
from
8e860d7
to
9990de4
Compare
vjardin
changed the title
Thanks to this definition, we'll be able to ave cleaner Makefiles.
the .so file should remain for the target when dynamic linkage is used. If the file is removed, none of the applications will be loadable.
start using DYNAMIC in order to be aligned with the --enable-all-static argument.
vjardin
force-pushed
the
fix_br_makefile
branch
from
2731a2c
to
cbc9640
Compare
DPDK assumes that the library shall be into /usr/lib instead of /usr/lib/mstflint/
|
Meanwhile, this pull request is included with the Buildroot package until it'll become mainstream. |
|
Thank you for your contribution. Below is some background information to clarify the concerns you're addressing with your patches. Mstflint is an open-source part of NVIDIA firmware tools (https://network.nvidia.com/products/adapter-software/firmware-tools/ - a proprietary code). Our typical process involves incorporating certain internal product features into the mstflint codebase with each release. While mstflint releases generally (but not always) align with mft releases, the mft undergoes a thorough testing cycle across various os-distribution-adapter combinations. For mstflint we only perform basic functionality testing. We do not deploy dynamic libraries for either mft or mstflint as well as we do not test neither of products in that configuration, all tools are statically linked with all its components. If you are aiming to link your code with the mtcr_ul library, please use its static version. Unfortunately, we cannot release something that has not been tested. Changing our internal process with this regard may take a little bit less than forever. Regarding your concern about the path used for placing the library: both mft and mstflint deploy its own versions of mtcr_ul library. To distinguish between them when both products are installed, we place the open-source version under /usr/lib/mstflint. To summarize: we can only consider the portion of your patch related to the hardcoded paths. Additionally, we do not work directly on the master branch. All development is done on the master_devel. For each patch from the community, we first absorb it to the internal product, subject it to our internal test cycle, and then it becomes a part of official release. This usually happens (though not always) once every three months. |
This serie is used to avoid hard coded path that do not work for any distributions. For instance, when buildroot is used or when cross compilation is used, the path of the headers cannot be enforce with /usr/include .