I wonder what are all possible criteria for Bitbake to decide to skip
recipe (please compare to reports delivered by bitbake-layers). Recipe
overlay, recipe version, others? Which knowledge source to find more
details in?

Maybe you can explain why you want to know why bitbake would skip a
recipe and that will help us to help you.

Thanks for input.
In meantime certain time ago, I guess there were different types of bitbake-layers output
(I has short opportunity to analyze) which comment several number of packages/recipes
by "recipe skipped" or something like this.
I like to be able to interpret that information properly, what's behind recipe skipped.

Regards
kd

On Tue, 4 May 2021, Quentin Schulz wrote:

On Tue, May 04, 2021 at 06:00:10AM -0400, Robert P. J. Day wrote:

On Tue, 4 May 2021, Quentin Schulz wrote:

Hi Robert,

On Tue, Apr 27, 2021 at 08:41:25AM -0400, Robert P. J. Day wrote:

for the first time, i'm digging around in the docs for how to
properly license various types of recipes, so a couple simple
questions to start with, at least so i can make a first pass of
cleaning up some content in front of me.

as we established recently, packagegroup files need no
licensing, the obvious observation being that they represent the
collection of licenses that comprise them. however, i notice that
the packagegroup.bbclass file does indeed define a default
license:

LICENSE ?= "MIT"

so not only does a packagegroup have a default (MIT) license, but
it's conditional suggesting one could give it a different license.
what other licenses would make sense for a packagegroup? I'm
sticking with the default that packagegroup recipe files need no
LICENSE assignment, but now i'm curious as to what other options
there are (or perhaps that that default assignment in
packagegroup.bbclass is obsolete).

Wild guess: all packages need a license. MIT is quite permissive so
safe as a default?

superficially makes sense, except that a packagegroup does not
really define a "package". perhaps all *recipe* files need a license

They do define packages. Empty packages, but still packages. Look into
deploy/ipk and search for *packagegroup*, you'll see some.

It's probably a requirement/feature of package managers, so that you
install one package which has a dependency on many others, and the
latter are just pulled in by the package manager directly.

but, again, it's not clear how a packagegroup license should percolate
down to the packages it contains. or how things would percolate up.

It does not apply to packages it RDEPENDS on, it applies to the packages
created by the packagegroup recipe, each of them then RDEPENDS on other
packages (with potentially (and often) licensed differently).

ah, now it makes sense, thanks.

rday

On Tue, May 04, 2021 at 06:00:10AM -0400, Robert P. J. Day wrote:

On Tue, 4 May 2021, Quentin Schulz wrote:

Hi Robert,

On Tue, Apr 27, 2021 at 08:41:25AM -0400, Robert P. J. Day wrote:

for the first time, i'm digging around in the docs for how to
properly license various types of recipes, so a couple simple
questions to start with, at least so i can make a first pass of
cleaning up some content in front of me.

as we established recently, packagegroup files need no
licensing, the obvious observation being that they represent the
collection of licenses that comprise them. however, i notice that
the packagegroup.bbclass file does indeed define a default
license:

LICENSE ?= "MIT"

so not only does a packagegroup have a default (MIT) license, but
it's conditional suggesting one could give it a different license.
what other licenses would make sense for a packagegroup? I'm
sticking with the default that packagegroup recipe files need no
LICENSE assignment, but now i'm curious as to what other options
there are (or perhaps that that default assignment in
packagegroup.bbclass is obsolete).

Wild guess: all packages need a license. MIT is quite permissive so
safe as a default?

superficially makes sense, except that a packagegroup does not
really define a "package". perhaps all *recipe* files need a license

They do define packages. Empty packages, but still packages. Look into
deploy/ipk and search for *packagegroup*, you'll see some.

It's probably a requirement/feature of package managers, so that you
install one package which has a dependency on many others, and the
latter are just pulled in by the package manager directly.

but, again, it's not clear how a packagegroup license should percolate
down to the packages it contains. or how things would percolate up.

It does not apply to packages it RDEPENDS on, it applies to the packages
created by the packagegroup recipe, each of them then RDEPENDS on other
packages (with potentially (and often) licensed differently).

Cheers,
Quentin

On Tue, 4 May 2021, Quentin Schulz wrote:

Hi Robert,

On Tue, Apr 27, 2021 at 08:41:25AM -0400, Robert P. J. Day wrote:

for the first time, i'm digging around in the docs for how to
properly license various types of recipes, so a couple simple
questions to start with, at least so i can make a first pass of
cleaning up some content in front of me.

as we established recently, packagegroup files need no
licensing, the obvious observation being that they represent the
collection of licenses that comprise them. however, i notice that
the packagegroup.bbclass file does indeed define a default
license:

LICENSE ?= "MIT"

so not only does a packagegroup have a default (MIT) license, but
it's conditional suggesting one could give it a different license.
what other licenses would make sense for a packagegroup? I'm
sticking with the default that packagegroup recipe files need no
LICENSE assignment, but now i'm curious as to what other options
there are (or perhaps that that default assignment in
packagegroup.bbclass is obsolete).

Wild guess: all packages need a license. MIT is quite permissive so
safe as a default?

superficially makes sense, except that a packagegroup does not
really define a "package". perhaps all *recipe* files need a license
but, again, it's not clear how a packagegroup license should percolate
down to the packages it contains. or how things would percolate up.

suddenly, i want some coffee.

the same sort of question can be asked about image files,
including the generic OE core-image*.bb recipe files. of all those
current core-image files:

core-image-base.bb
core-image-minimal.bb
core-image-minimal-dev.bb
core-image-minimal-initramfs.bb
core-image-minimal-mtdutils.bb
core-image-tiny-initramfs.bb

fail into two camps.

the first sets a license, then inherits core-image:

LICENSE = "MIT"
inherit core-image

the second type simply "require"s one of the other recipe files so it
has no need to set its own license, as in core-image-minimal-dev.bb:

require core-image-minimal.bb

similar to packagegroups, does a core-image recipe really need a
separate LICENSE setting, or could that be added to core-image.bbclass
to centralize it (if it's even needed at all)?

Don't know about this one but I guess it's some rest of the original
implementation where I guess everything needed a LICENSE?

I can only guess, but maybe this'll start a discussion :)

finally, WRT .bbappend files, the original recipes will have
their own licenses and if the .bbappend file is doing nothing but
adding some configuration (you know, PACKAGECONFIG, EXTRA_OEMAKE,
that sort of thing), then there should be no need for licensing in
the bbappend file. does all this sound reasonable so far?

I wouldn't expect any bbappend to modify a package/recipe license
without changing the sources (version bump). But not much experience
with that, so might be a valid use case?

i'm not *trying* to be overly pedantic (well, yeah, i am :-), but
given the importance of licensing these days, i want to understand how
they work far more than i do at the moment, especially since my new
co-workers are asking me about them.

rday

Hi Robert,

On Tue, Apr 27, 2021 at 08:41:25AM -0400, Robert P. J. Day wrote:

for the first time, i'm digging around in the docs for how to
properly license various types of recipes, so a couple simple
questions to start with, at least so i can make a first pass of
cleaning up some content in front of me.

as we established recently, packagegroup files need no licensing,
the obvious observation being that they represent the collection of
licenses that comprise them. however, i notice that the
packagegroup.bbclass file does indeed define a default license:

LICENSE ?= "MIT"

so not only does a packagegroup have a default (MIT) license, but it's
conditional suggesting one could give it a different license. what
other licenses would make sense for a packagegroup? I'm sticking with
the default that packagegroup recipe files need no LICENSE assignment,
but now i'm curious as to what other options there are (or perhaps
that that default assignment in packagegroup.bbclass is obsolete).

Wild guess: all packages need a license. MIT is quite permissive so safe
as a default?

the same sort of question can be asked about image files, including
the generic OE core-image*.bb recipe files. of all those current
core-image files:

core-image-base.bb
core-image-minimal.bb
core-image-minimal-dev.bb
core-image-minimal-initramfs.bb
core-image-minimal-mtdutils.bb
core-image-tiny-initramfs.bb

fail into two camps.

the first sets a license, then inherits core-image:

LICENSE = "MIT"
inherit core-image

the second type simply "require"s one of the other recipe files so it
has no need to set its own license, as in core-image-minimal-dev.bb:

require core-image-minimal.bb

similar to packagegroups, does a core-image recipe really need a
separate LICENSE setting, or could that be added to core-image.bbclass
to centralize it (if it's even needed at all)?

Don't know about this one but I guess it's some rest of the original
implementation where I guess everything needed a LICENSE?

I can only guess, but maybe this'll start a discussion :)

finally, WRT .bbappend files, the original recipes will have their
own licenses and if the .bbappend file is doing nothing but adding
some configuration (you know, PACKAGECONFIG, EXTRA_OEMAKE, that sort
of thing), then there should be no need for licensing in the bbappend
file. does all this sound reasonable so far?

I wouldn't expect any bbappend to modify a package/recipe license
without changing the sources (version bump). But not much experience
with that, so might be a valid use case?

Cheers,
Quentin

On Mon, May 03, 2021 at 11:25:09AM -0700, Sven via lists.yoctoproject.org wrote:

Hi,

I have put together a recipe inheriting from go-mod. This project depends on out-of-repo modules that sit in private repos. As long as the SSH key required to pull the requirements is present as a file (under $HOME/.ssh), everything works fine. However, as soon as the SSH credentials are only available via agent and $SSH_AUTH_SOCK, the do_compile step fails. I have traced this down to the fact that the $SSH_AUTH_SOCK environment variable is not available to do_compile which is when the requirements are pulled. This is the sort of error message I get:

On Ubuntu, when the password for my SSH key is in the gnome keyring, I
run:
eval $(/usr/bin/gnome-keyring-daemon --start --components=pkcs11,secrets,ssh); export SSH_AUTH_SOCK
in the terminal that will run bitbake.

No idea if it's the proper way to make it work, but for the few times I
need to fetch new revisions of private repos, it's been good enough to
me.

Hope this helps,
Cheers,
Quentin

On 2021-04-26 12:25 p.m., keydi wrote:

I wonder what are all possible criteria for Bitbake to decide to skip recipe (please compare to reports delivered by bitbake-layers). Recipe overlay, recipe version, others? Which knowledge source to find more details in?

I'm not aware of a specific list but two that come to mind are,
(in meta-openembedded):

$ rgrep PNBLACKLIST meta-*| wc -l
10
$ rgrep COMPATIBLE_HOST meta-*| wc -l
115


Maybe you can explain why you want to know why bitbake would skip a
recipe and that will help us to help you.



../Randy

--
# Randy MacLeod
# Wind River Linux

Hi,

I have put together a recipe inheriting from go-mod. This project depends on out-of-repo modules that sit in private repos. As long as the SSH key required to pull the requirements is present as a file (under $HOME/.ssh), everything works fine. However, as soon as the SSH credentials are only available via agent and $SSH_AUTH_SOCK, the do_compile step fails. I have traced this down to the fact that the $SSH_AUTH_SOCK environment variable is not available to do_compile which is when the requirements are pulled. This is the sort of error message I get:

ERROR: mypackage-git-r0 do_compile: Execution of '[...]/mypackage/git-r0/temp/run.do_compile.20076' failed with exit code 1:
# cd .; git ls-remote https://bitbucket.org/myorg/some-requirement
Permission denied (publickey).
fatal: Could not read from remote repository.

Note, that the do_fetch step succeeds in pulling the actual repo. I tried fixing the problem by wrapping the do_compile function and providing $SSH_AUTH_SOCK from the original environment:

def origenv(d, var):
return d.getVar("BB_ORIGENV", False).getVar(var, False)

do_compile() {
if [ -n "${@origenv(d, 'SSH_AUTH_SOCK') or ''}" ]; then
export SSH_AUTH_SOCK="${@origenv(d, 'SSH_AUTH_SOCK')}"
fi
go_do_compile
}

This allows the do_compile step (and all subsequent steps) to finish successfully. However, that way, I get a bunch of errors like this (cleansstate does not help):

ERROR: When reparsing [...]/mypackage_git.bb:do_compile, the basehash value changed from eb51e4ec321c723587cec03bb9b33b94ee43e0b0939eb43b52824e3d5cfebec2 to 2bb034f43856917d6454a56b32946b1c68cf7f286b20fd7a7eaf1bfd2a92d34f. The metadata is not deterministic and this needs to be fixed.
ERROR: The following commands may help:
ERROR: $ bitbake mypackage -cdo_compile -Snone
ERROR: Then:
ERROR: $ bitbake mypackage -cdo_compile -Sprintdiff

Neither command helps to fix this. What can I do? I'm on poky yocto-3.1.5-18-gbb7747497a.

Best regards,
Sven

On 5/2/21 11:41 PM, Martin Jansa wrote:

* works around:
http://errors.yoctoproject.org/Errors/Details/580076/
elfutils-0.148/src/ar.c:1088:56: error: '%-*ld' directive output may be truncated writing between 6 and 10 bytes into a region of size 7 [-Werror=format-truncation=]
1088 | snprintf (tmpbuf, sizeof (tmpbuf), ofmt ? "%-*lo" : "%-*ld", bufsize, val);
| ^~~~~
elfutils-0.148/src/ar.c:1088:55: note: directive argument in the range [0, 4294967295]
1088 | snprintf (tmpbuf, sizeof (tmpbuf), ofmt ? "%-*lo" : "%-*ld", bufsize, val);
| ^~~~~~~
http://errors.yoctoproject.org/Errors/Details/580078/
elfutils-0.148/tests/addrcfi.c:90:16: error: 'dwarf_frame_register' accessing 96 bytes in a region of size 64 [-Werror=stringop-overflow=]
90 | int result = dwarf_frame_register (stuff->frame, regno, ops_mem, &ops, &nops);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

looks ok. Thanks for fixing it

Signed-off-by: Martin Jansa <Martin.Jansa@...>
---
recipes-devtools/elfutils/elfutils_0.148.bb | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/recipes-devtools/elfutils/elfutils_0.148.bb b/recipes-devtools/elfutils/elfutils_0.148.bb
index 1f07a28..654a715 100644
--- a/recipes-devtools/elfutils/elfutils_0.148.bb
+++ b/recipes-devtools/elfutils/elfutils_0.148.bb
@@ -61,6 +61,14 @@ inherit autotools gettext
# but 0.175 has different license, so to be safe don't backport the fix, just ignore the issue
CFLAGS += "-Wno-error=missing-attributes"
+# There is a fix in 0.171 version (commit b10d7eb74064c5906f031cd17c0f82041c6a4ca1)
+# but 0.171 has different license, so to be safe don't backport the fix, just ignore the issue
+CFLAGS += "-Wno-error=format-truncation="
+
+# There is a fix in 0.182 version (commit 5621fe5443da23112170235dd5cac161e5c75e65)
+# but 0.182 has different license, so to be safe don't backport the fix, just ignore the issue
+CFLAGS += "-Wno-error=stringop-overflow="
+
EXTRA_OECONF = "--program-prefix=eu- --without-lzma"
EXTRA_OECONF_append_class-native = " --without-bzlib"
EXTRA_OECONF_append_libc-uclibc = " --enable-uclibc"

* works around:
http://errors.yoctoproject.org/Errors/Details/580076/
elfutils-0.148/src/ar.c:1088:56: error: '%-*ld' directive output may be truncated writing between 6 and 10 bytes into a region of size 7 [-Werror=format-truncation=]
1088 | snprintf (tmpbuf, sizeof (tmpbuf), ofmt ? "%-*lo" : "%-*ld", bufsize, val);
| ^~~~~
elfutils-0.148/src/ar.c:1088:55: note: directive argument in the range [0, 4294967295]
1088 | snprintf (tmpbuf, sizeof (tmpbuf), ofmt ? "%-*lo" : "%-*ld", bufsize, val);
| ^~~~~~~

http://errors.yoctoproject.org/Errors/Details/580078/
elfutils-0.148/tests/addrcfi.c:90:16: error: 'dwarf_frame_register' accessing 96 bytes in a region of size 64 [-Werror=stringop-overflow=]
90 | int result = dwarf_frame_register (stuff->frame, regno, ops_mem, &ops, &nops);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Martin Jansa <Martin.Jansa@...>
---
recipes-devtools/elfutils/elfutils_0.148.bb | 8 ++++++++
1 file changed, 8 insertions(+)

diff --git a/recipes-devtools/elfutils/elfutils_0.148.bb b/recipes-devtools/elfutils/elfutils_0.148.bb
index 1f07a28..654a715 100644
--- a/recipes-devtools/elfutils/elfutils_0.148.bb
+++ b/recipes-devtools/elfutils/elfutils_0.148.bb
@@ -61,6 +61,14 @@ inherit autotools gettext
# but 0.175 has different license, so to be safe don't backport the fix, just ignore the issue
CFLAGS += "-Wno-error=missing-attributes"

+# There is a fix in 0.171 version (commit b10d7eb74064c5906f031cd17c0f82041c6a4ca1)
+# but 0.171 has different license, so to be safe don't backport the fix, just ignore the issue
+CFLAGS += "-Wno-error=format-truncation="
+
+# There is a fix in 0.182 version (commit 5621fe5443da23112170235dd5cac161e5c75e65)
+# but 0.182 has different license, so to be safe don't backport the fix, just ignore the issue
+CFLAGS += "-Wno-error=stringop-overflow="
+
EXTRA_OECONF = "--program-prefix=eu- --without-lzma"
EXTRA_OECONF_append_class-native = " --without-bzlib"
EXTRA_OECONF_append_libc-uclibc = " --enable-uclibc"
--
2.30.2

I have been inactive for an extended period.

Signed-off-by: Philip Tricca <flihp@...>
---
MAINTAINERS | 5 -----
1 file changed, 5 deletions(-)

diff --git a/MAINTAINERS b/MAINTAINERS
index 36c451f..3166b32 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -24,11 +24,6 @@ F: conf
F: classes
F: recipes-*
=20
-M: Philip Tricca <flihp@...>
-F: conf
-F: classes
-F: recipes-*
-
COMMON
M: Yi Zhao <yi.zhao@...>
F: conf
--=20
2.20.1