Laszlo Ersek <ler...@redhat.com> writes:
> On 03/06/19 13:44, Daniel P. Berrangé wrote: >> On Wed, Mar 06, 2019 at 01:30:06PM +0100, Laszlo Ersek wrote: >>> Hi, >>> >>> in order to build any OVMF platform firmware image from the submodule at >>> "roms/edk2", the (recursive) OpenSSL submodule at >>> "roms/edk2/CryptoPkg/Library/OpensslLib/openssl" needs to be initialized >>> as well. >>> >>> Am I right to think this would be the first recursive submodule in QEMU? >>> How should I approach this? (I see we have a related script at >>> "scripts/git-submodule.sh".) >> >> The scripts/git-submodule.sh file is called automatically by "make" >> in QEMU to activate any submodules that are required during the normal >> build a QEMU developer does. >> >> The ROM submodules are special though. AFAIK, these are never built as a >> side effect of the QEMU build process, so never need to be initialized by >> the git-submodule.sh script. Developers always just use the pre-built >> ROM files bundled in QEMU. The only people checking out the ROM submodules >> are the maintainers who periodically build a new binary ROM. >> >> So in this sense the fact that EDK has submodules shouldn't be a factor, >> as we would not expect EDK to be built by regular QEMU developers > > Thank you for the explanation. I missed that the "configure" script > manipulated the "git_submodules" variable explicitly. (Which seems to be > the origin for GIT_SUBMODULES.) It covers only a subset of > ".gitmodules". Yes those are the ones we actually need to build QEMU, so libraries or test code needed for unit testing (in the softfloat case). -- Alex Bennée
