This series explores Daniels idea he proposed in [1], and is a successor to the v3.
I decided to go with the RFC tag because there are several things I have doubts about, and I hope somebody more familiar with the topic can help me out. I kept running into situations where meson felt like something I have to cheat and defeat. One thing that required special attention is keeping things functional from the build tree (i.e. not requiring an install). First I failed to generate a meson output form a custom_target() in a qemu target specific folder. So I decided to generate scripts that can be used both for installation and during compilation to massage the build tree so that goal is accomplished. The solution for running the scripts as a part of the build is kind of ugly, because for custom_target() output is mandatory, but my output is invalid. So I lied. When I thought, I was done with an RFC quality solution, I tested it from scratch (deleting my builddir), and it turns out for meson.add_install_script(script_name) script_name is supposed to be around, and executable already at configure time. So my generated scripts won't do because not around in the beginning. I decided to work that around quick and dirty by introducing a wrapper script. One alternative to this would be to have the logic in a single script in the source-tree, and just generate the list of modules for each target, which would be a parameter of this single script along with the target name. The downside of this is that we require 2 new in source tree scripts this way as well, but it is less flexible, because it can't be used for other generated scripts. Another thing I'm not really super satisfied with is this modules_restricted interface, but for my use case it does the job, and when something new emerges we can re-evaluate. But before I continue agonizing on these matters, I would like to get some feedback regarding, is this approach to the problem what the community wants (or is the whole symlinks idea controversial). [1] https://mail.gnu.org/archive/html/qemu-s390x/2021-03/msg00206.html Changelog v3 --> RFC: * switch to the symlinks approach (Daniel) * split s390x and common in separate patches (Eduardo) Halil Pasic (2): modules: introduce target specific modules hw/s390x: modularize virtio-gpu-ccw hw/s390x/meson.build | 8 +++++- include/hw/s390x/css.h | 7 ----- include/hw/s390x/s390_flic.h | 3 +++ include/qemu/module.h | 2 ++ meson.build | 43 +++++++++++++++++++++++++++++- roms/SLOF | 2 +- roms/opensbi | 2 +- scripts/call_generated_script.sh | 6 +++++ scripts/modules/target-symlinks.sh | 31 +++++++++++++++++++++ softmmu/runstate.c | 1 + target/s390x/cpu.h | 9 ++++--- util/module.c | 14 ++++++++-- 12 files changed, 112 insertions(+), 16 deletions(-) create mode 100755 scripts/call_generated_script.sh create mode 100755 scripts/modules/target-symlinks.sh base-commit: 2615a5e433aeb812c300d3a48e1a88e1303e2339 -- 2.25.1
