Re: [PATCH v2 2/2] dpkg: directly deploy via sstate cache

[Zhihang Wei <wzh@ilbers.de>]

On 7/23/26 12:16, MOESSBAUER, Felix wrote:
> On Tue, 2026-07-21 at 17:04 +0200, Felix Moessbauer wrote:
>> By that, we can get rid of the custom sstate logic, which should be
>> avoided if not needed. Further, the deploy process is now cleaner, as we
>> deploy from a dedicated dir instead of the WORKDIR.
>>
>> Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
>> ---
>> Changes since v1:
>>
>> - adapt barebox.bbclass to deploy from DEPLOYDIR instead of WORKDIR
>>
>>   doc/technical_overview.md             |  2 +-
>>   meta/classes-recipe/barebox.bbclass   |  4 +--
>>   meta/classes-recipe/dpkg-base.bbclass | 37 +++++++++------------------
>>   3 files changed, 15 insertions(+), 28 deletions(-)
>>
>> diff --git a/doc/technical_overview.md b/doc/technical_overview.md
>> index f7e00f7b..fe427cae 100644
>> --- a/doc/technical_overview.md
>> +++ b/doc/technical_overview.md
>> @@ -217,7 +217,7 @@ Both consist of the following steps:
>>      finally unmount again (`dpkg_undo_mounts`).
>>   
>>   6. Task `do_deploy_deb`: add successfully built packages
>> -   `${WORKDIR}/*.deb` to the isar-apt repository
>> +   `${DEPLOYDIR}/*.deb` to the isar-apt repository
>>      `${REPO_ISAR_DIR}/${DISTRO}`
>>   
>>   ## 3.6 Populate Target Filesystem
>> diff --git a/meta/classes-recipe/barebox.bbclass b/meta/classes-recipe/barebox.bbclass
>> index 45d77e3c..c9a767d5 100644
>> --- a/meta/classes-recipe/barebox.bbclass
>> +++ b/meta/classes-recipe/barebox.bbclass
>> @@ -47,12 +47,12 @@ BAREBOX_BASE_BIN ?= "barebox"
>>   
>>   do_deploy[dirs] = "${DEPLOY_DIR_IMAGE}"
>>   do_deploy() {
>> -    dpkg --fsys-tarfile ${WORKDIR}/${PN}_${CHANGELOG_V}_${DISTRO_ARCH}.deb | \
>> +    dpkg --fsys-tarfile ${DEPLOYDIR}/${PN}_${CHANGELOG_V}_${DISTRO_ARCH}.deb | \
>>           tar xOf - ./usr/lib/barebox/barebox.bin \
>>           > "${DEPLOY_DIR_IMAGE}/${BAREBOX_BASE_BIN}.img"
>>       ln -sf ${BAREBOX_BASE_BIN}.img ${DEPLOY_DIR_IMAGE}/barebox.bin
>>   
>> -    dpkg --fsys-tarfile ${WORKDIR}/${PN}_${CHANGELOG_V}_${DISTRO_ARCH}.deb | \
>> +    dpkg --fsys-tarfile ${DEPLOYDIR}/${PN}_${CHANGELOG_V}_${DISTRO_ARCH}.deb | \
>>           tar xOf - ./usr/lib/barebox/barebox.config \
>>           > "${DEPLOY_DIR_IMAGE}/${BAREBOX_BASE_BIN}.config"
>>       ln -sf ${BAREBOX_BASE_BIN}.config ${DEPLOY_DIR_IMAGE}/barebox.config
>> diff --git a/meta/classes-recipe/dpkg-base.bbclass b/meta/classes-recipe/dpkg-base.bbclass
>> index db912e90..e2ed4dfb 100644
>> --- a/meta/classes-recipe/dpkg-base.bbclass
>> +++ b/meta/classes-recipe/dpkg-base.bbclass
>> @@ -12,6 +12,8 @@ inherit repository
>>   inherit deb-dl-dir
>>   inherit essential
>>   
>> +DEPLOYDIR = "${WORKDIR}/deploy"
>> +
>>   DEPENDS ?= ""
>>   RPROVIDES ?= "${PROVIDES}"
>>   
>> @@ -186,6 +188,10 @@ dpkg_prepare_unshare_ccache() {
>>       setfacl -m u:${UNSHARE_SUBUID_BASE}:rwX -m u:${@int(d.getVar('UNSHARE_SUBUID_BASE')) + 999}:rwx "${CCACHE_DIR}"
>>   }
>>   
>> +dpkg_collect_debs() {
>> +    find ${WORKDIR} -maxdepth 1 -name "*.deb" -exec mv {} ${DEPLOYDIR} \;
>> +}
>> +
>>   python do_dpkg_build() {
>>       bb.build.exec_func('dpkg_chroot_prepare', d)
>>       try:
>> @@ -193,45 +199,26 @@ python do_dpkg_build() {
>>       finally:
>>           bb.build.exec_func('dpkg_chroot_finalize', d)
>>   }
>> +do_dpkg_build[cleandirs] = "${DEPLOYDIR}"
>> +do_dpkg_build[sstate-plaindirs] = "${DEPLOYDIR}"
> Hi,
>
> I'm currently playing around with using a hash equivalence server in
> the testsuite (which brings major speedups on repeated runs and also
> reduces the storage size needed). However, when do_dpkg_build is served
> from the sstate cache, that WORKDIR-internal plaindir is not guaranteed
> to be materialized (e.g. when the task is covered via hash equivalence
> without a setscene restore). do_deploy_deb then runs against an empty
> deploy dir, breaking the build.
>
> I already have fix for that which substitutes this patch. However,
> depending on the timeline (and that this patch works in the current
> mode), I'm wondering if I should send that as a replacement or as a
> later improvement, together with a ton of sstate fixes that currently
> remain unnoticed.
>
> Felix

Hi Felix,

I would say send it as a later improvement because we are about to merge
this patch.

Zhihang

-- 
You received this message because you are subscribed to the Google Groups "isar-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to isar-users+unsubscribe@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/isar-users/5d2fd657-ce85-4b4a-a062-d32afeb37897%40ilbers.de.