On Mon, 25 Sept 2023 at 17:45, Alex Bennée <alex.bennee(a)linaro.org> wrote:
Paolo Bonzini <pbonzini(a)redhat.com> writes:
> On 9/25/23 16:48, Alex Bennée wrote:
>> If we update configure we should make sure we regenerate all the
>> compiler details. We should also ensure those details are upto date
>> before building the TCG tests.
>> Signed-off-by: Alex Bennée <alex.bennee(a)linaro.org>
>> ---
>> configure | 2 ++
>> 1 file changed, 2 insertions(+)
>> diff --git a/configure b/configure
>> index e83872571d..a95e0f5767 100755
>> --- a/configure
>> +++ b/configure
>> @@ -1788,6 +1788,8 @@ for target in $target_list; do
>> echo "HOST_GDB_SUPPORTS_ARCH=y" >>
"$config_target_mak"
>> fi
>> + echo "$config_target_mak: configure" >>
Makefile.prereqs
>
> This in practice is not adding anything; if "configure" changes then
> Makefile's dependency on config-host.mak will trigger a configure
> rerun anyway.
>
> If you want to add it, you should also add it for other config-*.mak
> files. However, I'd remove this line and just change
>
> -# 1. ensure config-host.mak is up-to-date
> +# 1. ensure config-host.mak is up-to-date. All other config-*.mak
> +# files for subdirectories will be updated as well.
Peter ran into a mismatch between config-host.mak and
tests/tcg/foo/config-target.mak in his build system so it didn't get
picked up at one point.
I did, but looking at the timestamps on the two files, the
problem wasn't that one file got updated and not the other:
$ grep CONFIG_PLUGIN build/x86/config-host.h
#undef CONFIG_PLUGIN
$ grep CONFIG_PLUGIN build/x86/tests/tcg/config-host.mak
CONFIG_PLUGIN=y
e104462:jammy:qemu$ ls -l build/x86/config-host.mak
build/x86/tests/tcg/config-host.mak
-rw-r--r-- 1 petmay01 petmay01 549 Sep 22 16:38 build/x86/config-host.mak
-rw-r--r-- 1 petmay01 petmay01 159 Sep 22 16:38
build/x86/tests/tcg/config-host.mak
(both newer than 'configure' itself by about 10 days.)
-- PMM