On Tue, 1 Jun 2021 at 12:01, Marc-André Lureau
<marcandre.lureau@gmail.com> wrote:
>
> Hi Peter
>
> On Tue, Jun 1, 2021 at 1:17 PM Peter Maydell <peter.maydell@linaro.org> wrote:
>>
>> On Sat, 29 May 2021 at 19:55, <marcandre.lureau@redhat.com> wrote:
>> >
>> > From: Marc-André Lureau <marcandre.lureau@redhat.com>
>> >
>> > The following changes since commit 62c0ac5041e9130b041adfa13a41583d3c3ddd24:
>> >
>> > Merge remote-tracking branch 'remotes/rth-gitlab/tags/pull-tcg-20210526' into staging (2021-05-28 16:25:21 +0100)
>> >
>> > are available in the Git repository at:
>> >
>> > git@github.com:elmarco/qemu.git tags/libslirp-pull-request
>> >
>> > for you to fetch changes up to b060428091c758781acc4d42849accc036d3c816:
>> >
>> > build-sys: make libslirp a meson subproject (2021-05-29 22:52:37 +0400)
>> >
>> > ----------------------------------------------------------------
>> > Update libslirp & make it a subproject
>> >
>> > ----------------------------------------------------------------
>>
>> All hosts, odd warnings on checkout and running configure:
>>
>> warning: unable to rmdir 'slirp': Directory not empty
>
>
> This one is from git itself. It doesn't clean up old submodule locations, even though they are actually "clean". git submodule "(re)move" has its limits I guess.
Yeah, I guess we have to live with this one.
>> make: Entering directory '/home/pm/qemu/build/all'
>> config-host.mak is out-of-date, running configure
>> GIT ui/keycodemapdb meson tests/fp/berkeley-testfloat-3
>> tests/fp/berkeley-softfloat-3 dtc capstone slirp
>> warn: ignoring non-existent submodule slirp
>
>
> However, I don't get this when simply running make. Maybe you run make in parallel, and config-host.mak didn't have the time to regenerate with a new GIT_SUBMODULES.
>
> I wonder if we miss a dependency like "git-submodule-update: config-host.mak" ?
Something looks like it's still using an old list of submodules.
Yes, but I don't see how I could tell git-submodule-update until after config-host.mak is regenerated and read again.
Paolo, any idea?
It's a transient issue, similar to the git warning.
> Running configure before make should also prevent this from happening.
Incremental build needs to keep working.
Sure, but one-step warnings during incremental build are blockers?
>>
>> BSD VMs: error message just before launching the VM (though the VM did
>> seem to then launch OK):
>>
>> Found ninja-1.8.2 at /usr/bin/ninja
>> ninja: no work to do.
>> (GIT="git" "/home/peter.maydell/qemu-netbsd/scripts/git-submodule.sh"
>> update ui/keycodemapdb meson tests/fp/berkeley-testfloat-3 tests/fp/be
>> rkeley-softfloat-3 dtc capstone slirp)
>> warn: ignoring non-existent submodule slirp
>> /usr/bin/python3 -B /home/peter.maydell/qemu-netbsd/tests/vm/netbsd
>> --debug --jobs 8 --verbose --image
>> "/home/peter.maydell/.cache/qemu
>> -vm/images/netbsd.img" --snapshot --build-qemu
>> /home/peter.maydell/qemu-netbsd --
>> DEBUG:root:Creating archive
>> /home/peter.maydell/qemu-netbsd/build/vm-test-6kefrq76.tmp/data-f706c.tar
>> for src_dir dir: /home/peter.maydell/q
>> emu-netbsd
>> error: pathspec 'slirp' did not match any file(s) known to git.
Maybe this is something needing updating in the "create the archive"
script?
Correct, my bad. No idea why I couldn't reproduce this before..
I guess we should run scripts/archive-source.sh in CI.
>>
>> clang sanitizer build: link failure:
>> subprojects/libslirp/libslirp.so.0.3.0.p/src_arp_table.c.o: In
>> function `arp_table_add':
>> /home/petmay01/linaro/qemu-for-merges/build/clang/../../subprojects/libslirp/src/arp_table.c:51:
>> undefined reference to `__ubsan_handle_type_mismatch_v1'
>> /home/petmay01/linaro/qemu-for-merges/build/clang/../../subprojects/libslirp/src/arp_table.c:51:
>> undefined reference to `__ubsan_handle_type_mismatch_v1'
>> /home/petmay01/linaro/qemu-for-merges/build/clang/../../subprojects/libslirp/src/arp_table.c:51:
>> undefined reference to `__ubsan_handle_type_mismatch_v1'
>> /home/petmay01/linaro/qemu-for-merges/build/clang/../../subprojects/libslirp/src/arp_table.c:34:
>> undefined reference to `__ubsan_handle_type_mismatch_v1'
>> /home/petmay01/linaro/qemu-for-merges/build/clang/../../subprojects/libslirp/src/arp_table.c:34:
>> undefined reference to `__ubsan_handle_type_mismatch_v1'
>> (and lots more similar)
> I don't get this when running make vm-build-netbsd. What else am I missing?
This isn't NetBSD related, it's just a clang sanitizer build on Linux.
I am running configure with '--enable-sanitizers' --cc=clang --cxx=clang++ --host-cc=clang, I can't reproduce.
What's your distro? (or meson + clang versions)
>> OSX: linker warnings linking libslirp.0.dylib:
>>
>>
>> [34/1977] Linking target subprojects/libslirp/libslirp.0.dylib
>> ld: warning: dylib
>> (/usr/local/Cellar/glib/2.68.0/lib/libgthread-2.0.dylib) was built for
>> newer macOS version (10.15) than being linked (10.4)
>> ld: warning: dylib
>> (/usr/local/Cellar/glib/2.68.0/lib/libglib-2.0.dylib) was built for
>> newer macOS version (10.15) than being linked (10.4)
>> ld: warning: dylib (/usr/local/opt/gettext/lib/libintl.dylib) was
>> built for newer macOS version (10.14) than being linked (10.4)
>>
>
> This looks related to:
> https://gitlab.freedesktop.org/slirp/libslirp/-/commit/410e296a52fb274648f8ecf53561eaab4b33c52c
>
> It could be that we need to use the version information from glib (or from any libraries used).
>
> It looks safe to ignore although I re-opened:
> https://gitlab.freedesktop.org/slirp/libslirp/-/issues/36#note_940695
I'm not generally a fan of ignoring warnings. I would prefer it if
we understood why it was happening and how shared libraries are
supposed to be being built.
I reverted the change. MacOS build can override the macosx-version-min with CFLAGS.