On Tuesday 18 August 2015 10:36:05 Mariano Lopez wrote:
> On 08/18/2015 08:27 AM, Paul Eggleton wrote:
> > So aside from the error due to blank lines in the package list that I know
> > you're looking into, I did have a couple of comments:
> >
> > On Thursday 06 August 2015 12:49:05 mariano.lo...@linux.intel.com wrote:
> >> From: Mariano Lopez <mariano.lo...@linux.intel.com>
> >>
> >> This adds a new manifest file for the packages that were
> >> deployed but not included into rootfs. The new manifest file
> >> is in the same directory as the rootfs manifest but the name is
> >> image_license.manifest.
> >>
> >> It also creates the directory structure for the packages and
> >> add the license file in such directories.
> >>
> >> The bootloader is an example of such packages.
> >>
> >> [YOCTO #6772]
> >>
> >> Signed-off-by: Mariano Lopez <mariano.lo...@linux.intel.com>
> >> ---
> >>
> >> meta/classes/license.bbclass | 126
> >>
> >> ++++++++++++++++++++++++++++++++++++++----- 1 file changed, 113
> >> insertions(+), 13 deletions(-)
> >>
> >> diff --git a/meta/classes/license.bbclass b/meta/classes/license.bbclass
> >> index 32e172a..d4ad263 100644
> >> --- a/meta/classes/license.bbclass
> >> +++ b/meta/classes/license.bbclass
> >> @@ -26,18 +26,14 @@ python write_package_manifest() {
> >>
> >> }
> >>
> >> python license_create_manifest() {
> >>
> >> - import re
> >>
> >> import oe.packagedata
> >> from oe.rootfs import image_list_installed_packages
> >>
> >> - bad_licenses = (d.getVar("INCOMPATIBLE_LICENSE", True) or
> >> "").split()
> >> - bad_licenses = map(lambda l: canonical_license(d, l), bad_licenses)
> >> - bad_licenses = expand_wildcard_licenses(d, bad_licenses)
> >> -
> >>
> >> build_images_from_feeds = d.getVar('BUILD_IMAGES_FROM_FEEDS', True)
> >>
> >> if build_images_from_feeds == "1":
> >> return 0
> >>
> >> + # Files that are installed in the rootfs
> >>
> >> pkg_dic = {}
> >>
> >> for pkg in image_list_installed_packages(d).split("\n"):
> >> pkg_info = os.path.join(d.getVar('PKGDATA_DIR', True),
> >>
> >> @@ -49,8 +45,109 @@ python license_create_manifest() {
> >>
> >> pkg_lic_name = "LICENSE_" + pkg_name
> >> pkg_dic[pkg_name]["LICENSE"] =
> >> pkg_dic[pkg_name][pkg_lic_name]
> >>
> >> - license_manifest = os.path.join(d.getVar('LICENSE_DIRECTORY', True),
> >> + # Files that are part of the image but no installed in rootfs
> >> + imgpkg_dic = {}
> >> + for pkg in list_deployed_packages(d, pkg_dic.keys()).split("\n"):
> >> + pkg_info = os.path.join(d.getVar('PKGDATA_DIR', True),
> >> + 'runtime', pkg)
> >> + pkg_name = os.path.basename(pkg_info)
> >> + imgpkg_dic[pkg] = oe.packagedata.read_pkgdatafile(pkg_info)
> >> + if not "LICENSE" in imgpkg_dic[pkg].keys():
> >> + pkg_lic_name = "LICENSE_%s" % pkg
> >> + imgpkg_dic[pkg]["LICENSE"] = imgpkg_dic[pkg][pkg_lic_name]
> >> +
> >> + rootfs_license_manifest = os.path.join(d.getVar('LICENSE_DIRECTORY',
> >> True), d.getVar('IMAGE_NAME', True), 'license.manifest')
> >> + image_license_manifest = os.path.join(d.getVar('LICENSE_DIRECTORY',
> >> True),
> >> + d.getVar('IMAGE_NAME', True),
> >> 'image_license.manifest')
> >
> > Rather than image_license.manifest can we call this
> > "image_extra.manifest"?
> > Then it's just a little bit more clear what this file is for.
>
> Sure
>
> >> + write_license_files(d,
> >> rootfs_license_manifest, pkg_dic)
> >> + write_license_files(d, image_license_manifest, imgpkg_dic)
> >> +}
> >> +
> >> +
> >> +def list_deployed_packages(d, installed_packages):
> >> + """
> >> + Get all the deployed packages (not installed on rootfs) of an image
> >> +
> >> + This function needs the BB_TASKDEPDATA from do_rootfs.
> >> + This won't work outside do_rootfs.
> >> + """
> >> + import oe.packagedata
> >> +
> >> + packages = []
> >> + taskdepdata = d.getVar("BB_TASKDEPDATA", True)
> >> + all_depends = imagetypes_getdepends(d)
> >> + all_depends += " %s" % d.getVar("EXTRA_IMAGEDEPENDS", True)
> >> + for depend in all_depends.split():
> >> + # Get package name without task
> >> + depend = depend.split(":")[0]
> >> + if not depend.endswith("native"):
> >> + pkgs_file = os.path.join(
> >> + d.getVar('PKGDATA_DIR', True), depend)
> >> + # Search for the file on the installed packages with the
> >> + # same name as depend if not found fallback to search
> >> + # the provider for that depend.
> >> + if not os.path.isfile(pkgs_file):
> >> + pkgs_file = ""
> >> + for taskdep in taskdepdata.itervalues():
> >> + # The fifth field of BB_TASKDEPDATA is PROVIDES
> >> + if depend in taskdep[4]:
> >> + pkgs_file = os.path.join(
> >> + d.getVar('PKGDATA_DIR', True),
> >> + taskdep[0])
> >> + if os.path.isfile(pkgs_file):
> >> + break
> >> + else:
> >> + pkgs_file = ""
> >> + if pkgs_file:
> >> + pkgs_dep = oe.packagedata.read_pkgdatafile(pkgs_file)
> >> + # There is no need to duplicate license info for
> >> + # derivated packages or packages in the other
> >> + # license manifest
> >> + for pkg in pkgs_dep['PACKAGES'].split():
> >> + if (pkg.endswith("-dbg") or
> >> + pkg.endswith("-dev") or
> >> + pkg.endswith("-doc") or
> >> + pkg.endswith("-locale") or
> >> + pkg.endswith("-localedata") or
> >> + pkg.endswith("-staticdev") or
> >
> > This kind of hardcoded list rings alarm bells. Are we really looking at
> > the right thing here if we have to do this kind of filtering?
>
> When there is a virtual dependency in IMAGE_DEPENDS the
> pkgdata file can have several installed packages in there
> (an example is the kernel in raspberry pi), so we need to get
> the list of the packages that were installed, but also filter
> those that are redundant.
Looking at the code, I'm not quite understanding how packages even come into
it. You're looking at PROVIDES, not RPROVIDES, right?
Also I just noticed this:
> + # There is no need to add the license for
> + # packages that were not deployed
> + pkg_deploy = os.path.join(
> + d.getVar("TMPDIR", True), "work",
> + d.getVar("MULTIMACH_TARGET_SYS", True),
> + pkg_data["PN"], pkg_pf, "temp",
> "run.do_deploy")
> + if os.path.isfile(pkg_deploy):
> + packages.append(pkg)
This is not a good way of figuring out that the deploy task ran - the logs
aren't guaranteed to be there. As we discussed earlier I think the proper way
to do this is to look in sstate-control (more appropriately the path pointed
to by SSTATE_MANIFESTS) and check if a manifest file exists for the recipe that
has put files in DEPLOY_DIR_IMAGE.
Cheers,
Paul
--
Paul Eggleton
Intel Open Source Technology Centre
--
_______________________________________________
Openembedded-core mailing list
Openembedded-core@lists.openembedded.org
http://lists.openembedded.org/mailman/listinfo/openembedded-core
