meta: Fixes for update-alternatives.bbclass and gtk-icon-cache.bbclass
PATCH v2 1/2 update-alternatives.bbclass: run update-alternatives firstly in postinst script
In reply to Robert YangRecipes like postfix run command newaliases in postinst, but newaliases is installed as newaliases.postfix, it needs run update-alternatives to update it to newaliases, so there was an error when installed postinst on target.
Fixed:
$ opkg install postfix Configuring postfix. ///var/lib/opkg/info/postfix.postinst: line 4: newaliases: command not found
Run update-alternatives firstly will fix the problem. And handle busybox as an exception since it needs set basic tools such as sed command firstly, otherwise update-alternatives doesn't work.
--- meta/classes/update-alternatives.bbclass | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-)
meta/classes/update-alternatives.bbclass +11 -2
@@ -265,6 +265,7 @@ python populate_packages_updatealternatives () { if not update_alternatives_enabled(d): return + bpn = d.getVar('BPN') # Do actual update alternatives processing for pkg in (d.getVar('PACKAGES') or "").split(): # Create post install/removal scripts
@@ -284,8 +285,16 @@ python populate_packages_updatealternatives () { bb.note('adding update-alternatives calls to postinst/prerm for %s' % pkg) bb.note('%s' % alt_setup_links) - postinst = d.getVar('pkg_postinst_%s' % pkg) or '#!/bin/sh\n' - postinst += alt_setup_links + postinst = d.getVar('pkg_postinst_%s' % pkg) + if postinst: + # Busybox needs handle basic tools such as sed command + # firstly, otherwise, update-alternatives doesn't work. + if bpn == 'busybox': + postinst += alt_setup_links + else: + postinst = alt_setup_links + postinst + else: + postinst = '#!/bin/sh\n' + alt_setup_links d.setVar('pkg_postinst_%s' % pkg, postinst) bb.note('%s' % alt_remove_links)
Signature · 1 line
2.7.4
v2 cover meta: Fixes for update-alternatives.bbclass and gtk-icon-cache.bbclass
* V2
- The patch for update-alternatives.bbclass makes update-alternatives runs firstly before other postinst scripts, since other postinst may run the tool which is installed by the package itself (e.g., newaliases is installed as newaliases.postfix, and update-alternatives will update it to newaliases). But this broke busybox' on target install, since update-alternatives requires some basic tools such as sed, there was a neat way in busybox' postinst to fix this problem, the v2 makes busybox' postinst script run firstly before update-alternatives.
Testinfo:
MACHINE = "qemux86"PACKAGE_CLASSES = "package_ipk package_rpm" (or "package_rpm package_ipk") TESTIMAGE_AUTO = "1"
$ bitbake core-image-sato core-image-minimal
The tests are OK.
* V1
- Initial version
// Robert
The following changes since commit 4bb3e8f98e2bbfcb20f1b32d2b7674d1a6cb47be:
local.conf.sample.extended: remove redundant RUNTIMETARGET assignment (2019-06-24 17:32:37 +0100)
are available in the git repository at:
git://git.openembedded.org/openembedded-core-contrib rbt/post
http://cgit.openembedded.org//log/?h=rbt/post
Robert Yang (2):
update-alternatives.bbclass: run update-alternatives firstly in
postinst script
gtk-icon-cache.bbclass: Depends on gtk+3
meta/classes/gtk-icon-cache.bbclass | 14 ++++++++++---- meta/classes/update-alternatives.bbclass | 13 +++++++++++-- 2 files changed, 21 insertions(+), 6 deletions(-)
Signature · 1 line
2.7.4
PATCH v2 2/2 gtk-icon-cache.bbclass: Depends on gtk+3
The gtk-update-icon-cache is provided by gtk+3, gdk-pixbuf-query-loaders is provided by gdk-pixbuf, and gtk+3 depends on gdk-pixbuf, so depends on gtk+3 can fix the problems.
--- meta/classes/gtk-icon-cache.bbclass | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-)
meta/classes/gtk-icon-cache.bbclass +10 -4
@@ -4,6 +4,11 @@ DEPENDS +=" ${@['hicolor-icon-theme', '']['${BPN}' == 'hicolor-icon-theme']} gtk PACKAGE_WRITE_DEPS += "gtk+3-native gdk-pixbuf-native" +inherit distro_features_check +ANY_OF_DISTRO_FEATURES = "${GTK3DISTROFEATURES}" + +DEPENDS += "gtk+3" + gtk_icon_cache_postinst() { if [ "x$D" != "x" ]; then $INTERCEPT_DIR/postinst_intercept update_icon_cache ${PKG} \
@@ -45,10 +50,11 @@ python populate_packages_append () { if not os.path.exists(icon_dir): continue - bb.note("adding hicolor-icon-theme dependency to %s" % pkg) - rdepends = ' ' + d.getVar('MLPREFIX', False) + "hicolor-icon-theme" - d.appendVar('RDEPENDS_%s' % pkg, rdepends) - + for dep in ('hicolor-icon-theme', 'gtk+3'): + bb.note("Adding %s dependency to %s" % (dep, pkg)) + rdepends = ' ' + d.getVar('MLPREFIX', False) + dep + d.appendVar('RDEPENDS_%s' % pkg, rdepends) + bb.note("adding gtk-icon-cache postinst and postrm scripts to %s" % pkg) postinst = d.getVar('pkg_postinst_%s' % pkg)
Signature · 1 line
2.7.4
Re: [PATCH v2 1/2] update-alternatives.bbclass: run update-alternatives firstly in postinst script
In reply to Robert YangQuoted 39 lines · Robert Yang
Recipes like postfix run command newaliases in postinst, but newaliases is installed as newaliases.postfix, it needs run update-alternatives to update it to newaliases, so there was an error when installed postinst on target.
Fixed:
$ opkg install postfix Configuring postfix. ///var/lib/opkg/info/postfix.postinst: line 4: newaliases: command not foundRun update-alternatives firstly will fix the problem. And handle busybox as an exception since it needs set basic tools such as sed command firstly, otherwise update-alternatives doesn't work.
Signed-off-by Robert Yang <[email protected]>--- meta/classes/update-alternatives.bbclass | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-)meta/classes/update-alternatives.bbclass +4
-2@@ -265,6 +265,7 @@ python populate_packages_updatealternatives () { if not update_alternatives_enabled(d): return + bpn = d.getVar('BPN') # Do actual update alternatives processing for pkg in (d.getVar('PACKAGES') or "").split(): # Create post install/removal scripts@@ -284,8 +285,16 @@ python populate_packages_updatealternatives () { bb.note('adding update-alternatives calls to postinst/prerm for %s' % pkg) bb.note('%s' % alt_setup_links) - postinst = d.getVar('pkg_postinst_%s' % pkg) or '#!/bin/sh\n' - postinst += alt_setup_links + postinst = d.getVar('pkg_postinst_%s' % pkg) + if postinst: + # Busybox needs handle basic tools such as sed command
meta/classes/update-alternatives.bbclass python populate_packages_updatealternatives () { · line 300
Quoted 10 lines
meta/classes/update-alternatives.bbclass +7
-0+ # firstly, otherwise, update-alternatives doesn't work. + if bpn == 'busybox': + postinst += alt_setup_links + else: + postinst = alt_setup_links + postinst + else: + postinst = '#!/bin/sh\n' + alt_setup_links d.setVar('pkg_postinst_%s' % pkg, postinst) bb.note('%s' % alt_remove_links)
We don't write classes like this. We could just keep piling special cases one on top of the other and it becomes a horrible unmaintainable mess. I therefore don't want to see recipes "special cased" like this in core code and we need to find another, better way to handle this.
Cheers,
Richard
Re: [PATCH v2 1/2] update-alternatives.bbclass: run update-alternatives firstly in postinst script
Quoted 49 lines · Richard Purdie
Quoted 48 lines · Robert Yang
Recipes like postfix run command newaliases in postinst, but newaliases is installed as newaliases.postfix, it needs run update-alternatives to update it to newaliases, so there was an error when installed postinst on target.
Fixed:
$ opkg install postfix Configuring postfix. ///var/lib/opkg/info/postfix.postinst: line 4: newaliases: command not foundRun update-alternatives firstly will fix the problem. And handle busybox as an exception since it needs set basic tools such as sed command firstly, otherwise update-alternatives doesn't work.
Signed-off-by Robert Yang <[email protected]>--- meta/classes/update-alternatives.bbclass | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-)meta/classes/update-alternatives.bbclass +11
-2@@ -265,6 +265,7 @@ python populate_packages_updatealternatives () { if not update_alternatives_enabled(d): return + bpn = d.getVar('BPN') # Do actual update alternatives processing for pkg in (d.getVar('PACKAGES') or "").split(): # Create post install/removal scripts@@ -284,8 +285,16 @@ python populate_packages_updatealternatives () { bb.note('adding update-alternatives calls to postinst/prerm for %s' % pkg) bb.note('%s' % alt_setup_links) - postinst = d.getVar('pkg_postinst_%s' % pkg) or '#!/bin/sh\n' - postinst += alt_setup_links + postinst = d.getVar('pkg_postinst_%s' % pkg) + if postinst: + # Busybox needs handle basic tools such as sed command + # firstly, otherwise, update-alternatives doesn't work. + if bpn == 'busybox': + postinst += alt_setup_links + else: + postinst = alt_setup_links + postinst + else: + postinst = '#!/bin/sh\n' + alt_setup_links d.setVar('pkg_postinst_%s' % pkg, postinst)
Quoted 6 lines
meta/classes/update-alternatives.bbclass python populate_packages_updatealternatives () { · line 300
Quoted 1 line
meta/classes/update-alternatives.bbclass +0
-0bb.note('%s' % alt_remove_links)We don't write classes like this. We could just keep piling special cases one on top of the other and it becomes a horrible unmaintainable mess. I therefore don't want to see recipes "special cased" like this in core code and we need to find another, better way to handle this.
Yes, make sense, I was also worried about. My first fix was in busybox.inc, but looks a little tricky (Add a populate_packages_updatealternatives_append() in busybox.inc) to make it work. Please see the V3.
// Robert
Quoted 8 lines
Cheers,
Richard