No announcement yet.

Wayland Protocols 1.24 Released With Improvement To DMA-BUF Protocol For Multi-GPUs

  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    Originally posted by lb90 View Post

    That's false.
    Care to elaborate?

    It's not even possible to build tests separately from the main code, it's either all or nothing and recompiling only tests with different settings hasn't worked for any meson project I tried to build. it's all or nothing. Sure with a bit of sed or gcc wrapper hacks everything is possible, but I mean as naturally as with a make file?

    I even seached their bugtracker and this just doesn't seem to be supported.


    • #12

      You can change per-target arguments, after that what needs rebuilding is rebuilt, the rest remains untouched. As for high level targets like make exposes (make docs, make tests, etc.) well, no, they're not supported out of the box, but you can just create an alias target 'docs' and run meson compile docs. But nobody does that, unfortunately. There's certainly room for improvement in this area!



      • #13
        Originally posted by binarybanana View Post
        Have you ever used meson as an end-user or are you just parroting what you've heard from the hivemind?
        IME meson is an overcomplicated and at the same time inflexible abomination. You can't easily override broken behavior like building binaries/libraries and tests using different compiler flags. If you try it will just blow your build away and recompile everything. With make and frinds it's just "make CFLAGGS=$CFLAGS LDFLAGS=$LDFLAGS $etc tests" after the main build. Simple, easy and reliable. Cross-builds also tend to be broken more often with meson...
        Make is not a golden as you think. That only works if someone has not used

        CFLAGS =
        CFLAGS+= I_am_Not_Building also on the meson command line you have like c_arg for c cflag cpp_args for c++ cflags.

        Binarybanana the reality is meson is quite flexible problem is the documentation is horrible..
        Like I guess you would have never seen the --native-file and --cross-file options where you can set a stack of setting at one. Reliable typing many overrides on a command-line where you could typo them or instead loading a configuration file in ini format with them all in there.

        Cross builds with meson do have their own fair share of issues. Autotools cross builds were also their own levels of hell.

        Yes the meson man page does not even mention the existance of the --native-file or --cross-file documentation. The written online documention mostly miss the existance of --native-file.

        Reality is the core of meson is a lot more flexable. The documentation on meson will make you think that meson is lacking functionality that meson in fact has.


        • #14
          Thanks for pointing that out, unfortunately, if I understood that correctly, that means unless the project comes with explicit targets like docs, tests, etc. changing any options between targets is not easily possible, even if the meson core is able to handle it in theory. c-args and friends don't really help when those targets don't exist and meson will rebuild *everything*, even my hyper-optimized binaries I want to test with lest optimized tests (because often the actual binary/library works/tests fine, but simplistic test cases get mangled by advanced optimizations leading to test failures while less optimized tests linking against the optimized library go through, in which case the actual code being tested works as expected. I noticed this happening for many projects when experimenting. And yes the perf uplift is real, just look at Micheal's own tests on here.).

          Maybe it's possible to shoehore the necessary tasgets in even if the project didn't include them originally. Patching is easy as long as the meson file stays relatively stable.. I'll check out those cross compile options too, that looks like it might be really useful.

          Thanks for tang your time, maybe it's not as bad as I thought after all! That maket me think if a makefile target for meson wouldn't also be possible..


          • #15
            Originally posted by binarybanana View Post
            Thanks for tang your time, maybe it's not as bad as I thought after all! That maket me think if a makefile target for meson wouldn't also be possible..
            Yes it possible that meson could be used instead of autotools to make makefiles. No its not supported. Yes they provide a sample for cmake of course you would have been after gnumake.

            The issue of makefile/autotool combination not having enough targets was a problem back in the 1990s that I remember running into. Yes I do agree the young meson we are suffering through this problem again.