From: Jan Kiszka <jan.kiszka@siemens.com>
To: Claudius Heine <claudius.heine.ext@siemens.com>,
isar-users <isar-users@googlegroups.com>,
Claudius Heine <ch@denx.de>,
Henning Schild <henning.schild@siemens.com>,
Andreas Reichel <Andreas.Reichel@tngtech.com>
Subject: Re: [RFC] Interface for installing bootloaders
Date: Wed, 5 Sep 2018 10:05:01 +0200 [thread overview]
Message-ID: <61083627-3554-2d07-564e-3da2eeb63bed@siemens.com> (raw)
In-Reply-To: <aa79c471-6610-5cdf-31a2-572ac9046f8a@siemens.com>
On 2018-09-05 09:57, Claudius Heine wrote:
> On 2018-09-05 09:47, Jan Kiszka wrote:
>> On 2018-09-05 09:34, Claudius Heine wrote:
>>> Hi Jan,
>>>
>>> On 2018-09-05 08:08, Jan Kiszka wrote:
>>>> On 2018-09-04 11:27, Claudius Heine wrote:
>>>>> Hi Jan,
>>>>>
>>>>> On 2018-09-04 09:17, [ext] Jan Kiszka wrote:
>>>>>> Hi all,
>>>>>>
>>>>>> for installing a bootloader into an image, a number of packages
>>>>>> have to be installed into the buildchroot. These can be prebuilt
>>>>>> packages or packages we build via Isar. Currently, we only handle
>>>>>> the case of WIC requiring prebuilt Debian packages, and that in a
>>>>>> hacky way (no proper distro abstraction, unneeded installation of
>>>>>> unused dependencies). This proposal aims at providing a more
>>>>>> holistic solution:
>>>>>>
>>>>>> - Introduce new variables that express the two types of dependencies:
>>>>>> - BOOTLOADER_PREBUILT
>>>>>> - BOOTLOADER_CUSTOM
>>>>>>
>>>>>> So there is no "WIC" in these, any imaging recipe can use them.
>>>>>>
>>>>>> - Imaging recipes (e.g. via classes/wic-img.bbclass) should
>>>>>> - install all packages from BOOTLOADER_PREBUILT and
>>>>>> BOOTLOADER_CUSTOM
>>>>>> into the buildchroot
>>>>>> - DEPEND on all recipes in BOOTLOADER_CUSTOM
>>>>>>
>>>>>> - Define some common dependencies in conf/distro/${DISTRO}.conf:
>>>>>> - GRUB_PREBUILT
>>>>>> - GRUB_PREBUILT_append_${DISTRO_ARCH}
>>>>>> - SYSLINUX_PREBUILT
>>>>>> - ...
>>>>>>
>>>>>> Those variables can then be used to initialize
>>>>>> BOOTLOADER_PREBUILT as
>>>>>> needed and also replace those nasty BUILDCHROOT_PREINSTALL_WIC in
>>>>>> buildchroot-target.bb.
>>>>>>
>>>>>> - Set BOOTLOADER_PREBUILT or BOOTLOADER_CUSTOM in the
>>>>>> conf/machine/${MACHINE}.conf or some overriding conf file -
>>>>>> basically
>>>>>> the same one that selects IMAGE_TYPE and WKS_FILE.
>>>>>>
>>>>>> Anything I missed? Better suggestions? Eventually, when we can
>>>>>> derive bitbake recipe dependencies automatically from package
>>>>>> dependencies, PREBUILT and CUSTOM could be folded together. Right
>>>>>> now, we need that split.
>>>>>>
>>>>>> Jan
>>>>>
>>>>> I have a class that I use for the current project that allows
>>>>> installing additional packages into the buildchroot from within the
>>>>> image. I didn't had time to prepare patches for it for mainline
>>>>> isar. But maybe that goes into to right direction.
>>>>>
>>>>> I attached it.
>>>>
>>>> ...
>>>>
>>>>> BUILDCHROOT_EXTRA_INSTALL ??= ""
>>>>>
>>>>> MOUNT_LOCKFILE = "${BUILDCHROOT_DIR}/mount.lock"
>>>>>
>>>>> BUILDCHROOT_DIR ??= "${BUILDCHROOT_TARGET_DIR}"
>>>>> BUILDCHROOT_BUILD_DEP ??= "buildchroot-target:do_build"
>>>>>
>>>>> do_buildchroot_extra_install[depends] = "${BUILDCHROOT_BUILD_DEP}"
>>>>> do_buildchroot_extra_install[deptask] = "do_deploy_deb"
>>>>> do_buildchroot_extra_install[stamp-extra-info] =
>>>>> "${DISTRO}-${MACHINE}"
>>>>> do_buildchroot_extra_install() {
>>>>> PACKAGES="${@" ".join(d.getVar("BUILDCHROOT_EXTRA_INSTALL",
>>>>> True).split())}"
>>>>>
>>>>
>>>> Can you explain the magic dance in the line above? I don't see its
>>>> need yet.
>>>>
>>>>> if [ -z "$PACKAGES" ]; then
>>>>
>>>> ...provided you do "[ -z $VAR ]" here to catch VAR=" ".
>>>
>>> In sh i get:
>>>
>>> $ tets="asdf asdf asdf asdf"
>>> $ [ -z $tets ] && echo yes
>>> sh: 2: [: asdf: unexpected operator
>>
>> Ah, that's what I missed. But then we want to .strip() the variable
>> for testing, rather than split/rejoin.
>
> Might work, but I don't like having possible '\n' and '\t' in between
> package names. (Ok, '\n' might not be an issue, since I think bitbake
> already escapes that.)
>
> For me that is the normal code snippet to convert bitbake 'arrays' into
> shell strings.
I find this pattern in pure form in exactly one poky recipe. Therefore
I'm not sure we need it in order to make a list built under bitbake in
various ways available as parameter string for a shell command. Let's see...
Jan
--
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux
next prev parent reply other threads:[~2018-09-05 8:05 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-09-04 7:17 Jan Kiszka
2018-09-04 9:27 ` Claudius Heine
2018-09-04 9:33 ` Jan Kiszka
2018-09-05 6:08 ` Jan Kiszka
2018-09-05 7:34 ` Claudius Heine
2018-09-05 7:47 ` Jan Kiszka
2018-09-05 7:57 ` Claudius Heine
2018-09-05 8:05 ` Jan Kiszka [this message]
2018-09-05 8:16 ` Claudius Heine
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=61083627-3554-2d07-564e-3da2eeb63bed@siemens.com \
--to=jan.kiszka@siemens.com \
--cc=Andreas.Reichel@tngtech.com \
--cc=ch@denx.de \
--cc=claudius.heine.ext@siemens.com \
--cc=henning.schild@siemens.com \
--cc=isar-users@googlegroups.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox