On 3/28/19 12:45 AM, Xi Ruoyao via lfs-dev wrote:
On 2019-03-28 15:15 +1000, James B via lfs-dev wrote:
On Wed, 27 Mar 2019 21:59:38 -0500
Bruce Dubbs via lfs-dev <[email protected]> wrote:
There are just too many problems with meson-0.50.0. Therefore I have
reverted to meson-0.49.2 in lfs-svn until it gets fixed.
What I don't understand is the mass exodus from a proven tools like autotools
(or cmake) to an immature, incomplete, and dependent-laden tools like meson,
to build critical system components :(
I don't think autotools are proven - I remember many notes about test failures
in autoconf and automake in LFS book. There are much more XFAILS.
The worst thing in autotools is the stupid libtool with a lot of .la files.
I agree with that. autotools was created for packages that want to be
available for many versions if Unix (solaris, hp-ux, etc) as well as
Linux. cmake and meson generally are targeted more specifically at
Linux where .la files are not needed.
For cmake, I always find it difficult to use. There is not even a consistent
way to specify the library install directory. Personally I classify cmake as an
example of "programs which are hard to integrate into the system, coming from
the mathematical guys". These math guys just throw each package into a
designated prefix, pullate environments like $PATH and $LD_LIBRARY_PATH (yes,
many guys in my lab don't even know ld.so.conf) and they believe everyone will
do that. When I build packages outside LFS/BLFS but with only cmake building
system avaliable, I just feel headache and want to say colorful metaphors.
I don't have a lot of problems with cmake any more. Generally when
checking a package using cmake I do:
mkdir build
cd build
cmake -L ..
that usually lists all the options a package understands and what
dependencies it finds. It's not always perfect, but usually is enough
to do what is needed.
-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
