This updated the port to the tip of blitz branch.
Details
poudriere testport passes on amd64 CURRENT.
Diff Detail
- Repository
- R11 FreeBSD ports repository
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
Why do we need the patches? I don't understand that since it builds for me w/o them, so I'd like to understand that.
- emulators/qemu-user-static-devel/files/patch-configure:
The port passes --python=${PYTHON_CMD} to the configure script and this makes it to die with error. The check looks like a safety belt to make sure that Python passed in --python and Python used by Meson are the same. This is true anyways for Ports tree.
- emulators/qemu-user-static-devel/files/patch-tests_fp_meson.build
The tests/fp/meson.build files tries to include a directory that comes from git submodule. This directory isn't present when pulling distfile from GitHub. I can arrange the port to pull this directory too, but all this looks to be test-related, so I thought it is safe to patch out.
- emulators/qemu-user-static-devel/files/patch-accel_tcg_user-exec.c
The compiler complained about undefined symbol T_PAGEFLT, which seems to be defined in x86/trap.h.
Sorry, I've been taking some time off and am just getting back into the swing of things
OK
- emulators/qemu-user-static-devel/files/patch-tests_fp_meson.build
The tests/fp/meson.build files tries to include a directory that comes from git submodule. This directory isn't present when pulling distfile from GitHub. I can arrange the port to pull this directory too, but all this looks to be test-related, so I thought it is safe to patch out.
OK. This is fine.
- emulators/qemu-user-static-devel/files/patch-accel_tcg_user-exec.c
The compiler complained about undefined symbol T_PAGEFLT, which seems to be defined in x86/trap.h.
Ah, I've fixed this upstream. I need to rebase the blitz branch to pick it up.
OK. I'd like to update the blitz branch this week. If I do, I'd like this to be rebased to that.
If I don't get to it this week, this looks good to go as it is.
emulators/qemu-user-static-devel/Makefile | ||
---|---|---|
50 | Cool that you've hooked this up. | |
emulators/qemu-user-static-devel/files/patch-tests_fp_meson.build | ||
10 | I'd add a comment here # disable fp tests since they live in a subrepo not fetched in the github tarball |
I've updated the 'next' blitz branch to 6.1.0rc1. You can drop the patch-accel_tcg_user-exec.c since that's now fully resolved and upstreamed.
Otherwise, this should be the same (modulo fluff).
So blitz-next will float over the next couple of weeks until 6.1.0 is done. I'll then make it the real blitz branch.
So other than a quibble over the config args, I think this is looking really good.
emulators/qemu-user-static-devel/Makefile | ||
---|---|---|
26–27 | This isn't directly used by bsd-user. Do we need it still? If so, that's cool, but if not we should remove it here. | |
54 | I know once upon a time we needed the long list of disable. These days I think we just need '--disable-system'. The rest is automatically excluded. Inside ports, we may still need the extra stuff. I needed to use clang11, since clang12.0.1 has issues (though the 12.0.1rc2 didn't). |
one last thing I forgot to mention here: clang 12.0.1 miscompiles bsd-user. 12.0.1rc2 is fine on a different machine, but I had to configure clang11 or gcc10 to build my development copies of bsd-user
emulators/qemu-user-static-devel/Makefile | ||
---|---|---|
26–27 |
It is still the case, the configure script fails if this dir isn't present. I can try patching it out, though. |
emulators/qemu-user-static-devel/Makefile | ||
---|---|---|
26–27 | If we still need it, better to just grab it and use upstream with fewer modifications. |
Looking good. Feel free to commit any time. And thanks for doing this and putting up with my flakiness giving you feedback.
This comment applies only to amd64 now. riscv and arm64 allocate trampolines at probe creation time. If that fails, an error can be signaled to libdtrace, and the user will get a nice error message.
In other words, I think there is no need to log anything on !amd64 if trampoline allocation fails.