[meta-networking][PATCH] networkmanager: Update to 1.18.0

[meta-networking][PATCH] networkmanager: Update to 1.18.0

[Andrei Gherzan]

One patch was dropped (0003-dlopen-failure.patch) as a similar one was
merged upstream. See:
https://github.com/NetworkManager/NetworkManager/commit/bd4957fcd78a0e64d26459e94ab3b84c91cd6f5f

Signed-off-by: Andrei Gherzan <andrei@gherzan.ro>
---
 .../networkmanager/0003-dlopen-failure.patch  | 34 -------------------
 ...ger_1.16.0.bb => networkmanager_1.18.0.bb} |  6 ++--
 2 files changed, 2 insertions(+), 38 deletions(-)
 delete mode 100644 meta-networking/recipes-connectivity/networkmanager/networkmanager/0003-dlopen-failure.patch
 rename meta-networking/recipes-connectivity/networkmanager/{networkmanager_1.16.0.bb => networkmanager_1.18.0.bb} (96%)

diff --git a/meta-networking/recipes-connectivity/networkmanager/networkmanager/0003-dlopen-failure.patch b/meta-networking/recipes-connectivity/networkmanager/networkmanager/0003-dlopen-failure.patch
deleted file mode 100644
index e99c22710..000000000
--- a/meta-networking/recipes-connectivity/networkmanager/networkmanager/0003-dlopen-failure.patch
+++ /dev/null
@@ -1,34 +0,0 @@
-From 57239fda56b68a8f3e413f7b6af5290ba0d86636 Mon Sep 17 00:00:00 2001
-From: =?UTF-8?q?Andreas=20M=C3=BCller?= <schnitzeltony@gmail.com>
-Date: Thu, 22 Mar 2018 18:18:06 +0100
-Subject: [PATCH] musl: dlopen is included so LD_LIBS="" instead of
- LD_LIBS="none required"
-MIME-Version: 1.0
-Content-Type: text/plain; charset=UTF-8
-Content-Transfer-Encoding: 8bit
-
-Upstream-Status: Pending
-
-Stolen from [1] and prettyfied slightly
-
-[1] https://github.com/voidlinux/void-packages/tree/master/srcpkgs/NetworkManager/patches
-
-Signed-off-by: Andreas Müller <schnitzeltony@gmail.com>
----
- configure.ac | 1 +
- 1 file changed, 1 insertion(+)
-
-diff --git a/configure.ac b/configure.ac
-index 487a266..96ae4f7 100644
---- a/configure.ac
-+++ b/configure.ac
-@@ -235,6 +235,7 @@ dnl
- dnl Checks for libdl - on certain platforms its part of libc
- dnl
- AC_SEARCH_LIBS([dlopen], [dl dld], [], [ac_cv_search_dlopen=])
-+AS_IF([test "$ac_cv_search_dlopen" = "none required"],[ac_cv_search_dlopen=""])
- AC_SUBST([DL_LIBS], "$ac_cv_search_dlopen")
- 
- PKG_CHECK_MODULES(GLIB, [gio-unix-2.0 >= 2.37.6 gmodule-2.0],
--- 
-2.14.3
diff --git a/meta-networking/recipes-connectivity/networkmanager/networkmanager_1.16.0.bb b/meta-networking/recipes-connectivity/networkmanager/networkmanager_1.18.0.bb
similarity index 96%
rename from meta-networking/recipes-connectivity/networkmanager/networkmanager_1.16.0.bb
rename to meta-networking/recipes-connectivity/networkmanager/networkmanager_1.18.0.bb
index 9f01f1bc1..9215a8676 100644
--- a/meta-networking/recipes-connectivity/networkmanager/networkmanager_1.16.0.bb
+++ b/meta-networking/recipes-connectivity/networkmanager/networkmanager_1.18.0.bb
@@ -26,15 +26,13 @@ SRC_URI = " \
     ${GNOME_MIRROR}/NetworkManager/${@gnome_verdir("${PV}")}/NetworkManager-${PV}.tar.xz \
     file://0001-Fixed-configure.ac-Fix-pkgconfig-sysroot-locations.patch \
     file://0002-Do-not-create-settings-settings-property-documentati.patch \
-    file://0003-dlopen-failure.patch \
 "
 SRC_URI_append_libc-musl = " \
     file://musl/0001-Fix-build-with-musl-systemd-specific.patch \
     file://musl/0002-Fix-build-with-musl.patch \
 "
-
-SRC_URI[md5sum] = "10abacaafb162a67d2942adf03e7e9e4"
-SRC_URI[sha256sum] = "8e962833b6ca03edda1bc57ed6614a7b8c2339531b44acef098d05f2324c5d2c"
+SRC_URI[md5sum] = "c8c27116a8083bab5d5fcca0d03e988a"
+SRC_URI[sha256sum] = "c6e8df25e5a3c7309bc17664be8971689314884cdd08afdd6b0847d29d2a8ba6"
 
 UPSTREAM_CHECK_URI = "${GNOME_MIRROR}/NetworkManager/1.16/"
 UPSTREAM_CHECK_REGEX = "NetworkManager\-(?P<pver>1\.10(\.\d+)+).tar.xz"
-- 
2.17.1

Re: [meta-networking][PATCH] networkmanager: Update to 1.18.0

[Khem Raj]

Hi Andrei

The musl patches need to be forward ported as well. They dont apply
cleanly anymore on top of this update.

On Wed, May 22, 2019 at 6:14 AM Andrei Gherzan <andrei@gherzan.ro> wrote:
>
> One patch was dropped (0003-dlopen-failure.patch) as a similar one was
> merged upstream. See:
> https://github.com/NetworkManager/NetworkManager/commit/bd4957fcd78a0e64d26459e94ab3b84c91cd6f5f
>
> Signed-off-by: Andrei Gherzan <andrei@gherzan.ro>
> ---
>  .../networkmanager/0003-dlopen-failure.patch  | 34 -------------------
>  ...ger_1.16.0.bb => networkmanager_1.18.0.bb} |  6 ++--
>  2 files changed, 2 insertions(+), 38 deletions(-)
>  delete mode 100644 meta-networking/recipes-connectivity/networkmanager/networkmanager/0003-dlopen-failure.patch
>  rename meta-networking/recipes-connectivity/networkmanager/{networkmanager_1.16.0.bb => networkmanager_1.18.0.bb} (96%)
>
> diff --git a/meta-networking/recipes-connectivity/networkmanager/networkmanager/0003-dlopen-failure.patch b/meta-networking/recipes-connectivity/networkmanager/networkmanager/0003-dlopen-failure.patch
> deleted file mode 100644
> index e99c22710..000000000
> --- a/meta-networking/recipes-connectivity/networkmanager/networkmanager/0003-dlopen-failure.patch
> +++ /dev/null
> @@ -1,34 +0,0 @@
> -From 57239fda56b68a8f3e413f7b6af5290ba0d86636 Mon Sep 17 00:00:00 2001
> -From: =?UTF-8?q?Andreas=20M=C3=BCller?= <schnitzeltony@gmail.com>
> -Date: Thu, 22 Mar 2018 18:18:06 +0100
> -Subject: [PATCH] musl: dlopen is included so LD_LIBS="" instead of
> - LD_LIBS="none required"
> -MIME-Version: 1.0
> -Content-Type: text/plain; charset=UTF-8
> -Content-Transfer-Encoding: 8bit
> -
> -Upstream-Status: Pending
> -
> -Stolen from [1] and prettyfied slightly
> -
> -[1] https://github.com/voidlinux/void-packages/tree/master/srcpkgs/NetworkManager/patches
> -
> -Signed-off-by: Andreas Müller <schnitzeltony@gmail.com>
> ----
> - configure.ac | 1 +
> - 1 file changed, 1 insertion(+)
> -
> -diff --git a/configure.ac b/configure.ac
> -index 487a266..96ae4f7 100644
> ---- a/configure.ac
> -+++ b/configure.ac
> -@@ -235,6 +235,7 @@ dnl
> - dnl Checks for libdl - on certain platforms its part of libc
> - dnl
> - AC_SEARCH_LIBS([dlopen], [dl dld], [], [ac_cv_search_dlopen=])
> -+AS_IF([test "$ac_cv_search_dlopen" = "none required"],[ac_cv_search_dlopen=""])
> - AC_SUBST([DL_LIBS], "$ac_cv_search_dlopen")
> -
> - PKG_CHECK_MODULES(GLIB, [gio-unix-2.0 >= 2.37.6 gmodule-2.0],
> ---
> -2.14.3
> diff --git a/meta-networking/recipes-connectivity/networkmanager/networkmanager_1.16.0.bb b/meta-networking/recipes-connectivity/networkmanager/networkmanager_1.18.0.bb
> similarity index 96%
> rename from meta-networking/recipes-connectivity/networkmanager/networkmanager_1.16.0.bb
> rename to meta-networking/recipes-connectivity/networkmanager/networkmanager_1.18.0.bb
> index 9f01f1bc1..9215a8676 100644
> --- a/meta-networking/recipes-connectivity/networkmanager/networkmanager_1.16.0.bb
> +++ b/meta-networking/recipes-connectivity/networkmanager/networkmanager_1.18.0.bb
> @@ -26,15 +26,13 @@ SRC_URI = " \
>      ${GNOME_MIRROR}/NetworkManager/${@gnome_verdir("${PV}")}/NetworkManager-${PV}.tar.xz \
>      file://0001-Fixed-configure.ac-Fix-pkgconfig-sysroot-locations.patch \
>      file://0002-Do-not-create-settings-settings-property-documentati.patch \
> -    file://0003-dlopen-failure.patch \
>  "
>  SRC_URI_append_libc-musl = " \
>      file://musl/0001-Fix-build-with-musl-systemd-specific.patch \
>      file://musl/0002-Fix-build-with-musl.patch \
>  "
> -
> -SRC_URI[md5sum] = "10abacaafb162a67d2942adf03e7e9e4"
> -SRC_URI[sha256sum] = "8e962833b6ca03edda1bc57ed6614a7b8c2339531b44acef098d05f2324c5d2c"
> +SRC_URI[md5sum] = "c8c27116a8083bab5d5fcca0d03e988a"
> +SRC_URI[sha256sum] = "c6e8df25e5a3c7309bc17664be8971689314884cdd08afdd6b0847d29d2a8ba6"
>
>  UPSTREAM_CHECK_URI = "${GNOME_MIRROR}/NetworkManager/1.16/"
>  UPSTREAM_CHECK_REGEX = "NetworkManager\-(?P<pver>1\.10(\.\d+)+).tar.xz"
> --
> 2.17.1
>
> --
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-devel

Re: [meta-networking][PATCH] networkmanager: Update to 1.18.0

[Andrei Gherzan]

On 28/05/2019 16.13, Khem Raj wrote:
> Hi Andrei
>
> The musl patches need to be forward ported as well. They dont apply
> cleanly anymore on top of this update.

Seems like. So all the contributions now need to go through local musl
and glibc testing? This sounds a little bit of an overhead and might
discourage people in contributing because usually everyone is working on
one libc variant. Just raising it as a concern. As per this specific
case, I'm going to fix it and push a V2.

-- 
Andrei Gherzan
gpg: rsa4096/D4D94F67AD0E9640 | t: @agherzan

Re: [meta-networking][PATCH] networkmanager: Update to 1.18.0

[Khem Raj]

On Tue, May 28, 2019 at 8:37 AM Andrei Gherzan <andrei@gherzan.ro> wrote:

> On 28/05/2019 16.13, Khem Raj wrote:
> > Hi Andrei
> >
> > The musl patches need to be forward ported as well. They dont apply
> > cleanly anymore on top of this update.
>
> Seems like. So all the contributions now need to go through local musl
> and glibc testing? This sounds a little bit of an overhead and might
> discourage people in contributing because usually everyone is working on
> one libc variant. Just raising it as a concern. As per this specific
> case, I'm going to fix it and push a V2.


I understand that and in some cases it’s even narrower where one might just
be working on one distribution alone for single architecture for that case
failure on architectures they don’t care is
also a nuisance

It’s wide and far testing we can achieve is good for end user same goes for
multiple architectures if someone tests a given architecture and reports
build  failures then it’s good to acknowledge that imo

It’s not a blocker I expect someone interested in a given regression of
this kind to chime and help if possible so don’t see it as an impediment

If we remain cognizant of the fact that upstream might have wider usecases
we will generally be better off cumulatively


>
> --
> Andrei Gherzan
> gpg: rsa4096/D4D94F67AD0E9640 | t: @agherzan
>
>

Re: [meta-networking][PATCH] networkmanager: Update to 1.18.0

[Adrian Bunk]

On Tue, May 28, 2019 at 09:14:17AM -0700, Khem Raj wrote:
> On Tue, May 28, 2019 at 8:37 AM Andrei Gherzan <andrei@gherzan.ro> wrote:
> 
> > On 28/05/2019 16.13, Khem Raj wrote:
> > > Hi Andrei
> > >
> > > The musl patches need to be forward ported as well. They dont apply
> > > cleanly anymore on top of this update.
> >
> > Seems like. So all the contributions now need to go through local musl
> > and glibc testing? This sounds a little bit of an overhead and might
> > discourage people in contributing because usually everyone is working on
> > one libc variant. Just raising it as a concern. As per this specific
> > case, I'm going to fix it and push a V2.
> 
> 
> I understand that and in some cases it’s even narrower where one might just
> be working on one distribution alone for single architecture for that case
> failure on architectures they don’t care is
> also a nuisance
> 
> It’s wide and far testing we can achieve is good for end user same goes for
> multiple architectures if someone tests a given architecture and reports
> build  failures then it’s good to acknowledge that imo
> 
> It’s not a blocker I expect someone interested in a given regression of
> this kind to chime and help if possible so don’t see it as an impediment

Is anyone actually using networkmanager with musl?

There are no obvious usecases where it makes sense, and OE seems to be
adding security vulnerabilities in the musl build that are not in the 
glibc build (e.g. -Dsecure_getenv=getenv) so users should better not
use it in any case.

> If we remain cognizant of the fact that upstream might have wider usecases
> we will generally be better off cumulatively

Upstream ships a copy of systemd code for use as library.

This already implies that trying to support it for musl will be a 
long-term pain of not upstreamable patches for OE contributors.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: [meta-networking][PATCH] networkmanager: Update to 1.18.0

[Andrei Gherzan]

Hi,

On 28/05/2019 19.06, Adrian Bunk wrote:
> On Tue, May 28, 2019 at 09:14:17AM -0700, Khem Raj wrote:
>> On Tue, May 28, 2019 at 8:37 AM Andrei Gherzan <andrei@gherzan.ro> wrote:
>>
>>> On 28/05/2019 16.13, Khem Raj wrote:
>>>> Hi Andrei
>>>>
>>>> The musl patches need to be forward ported as well. They dont apply
>>>> cleanly anymore on top of this update.
>>> Seems like. So all the contributions now need to go through local musl
>>> and glibc testing? This sounds a little bit of an overhead and might
>>> discourage people in contributing because usually everyone is working on
>>> one libc variant. Just raising it as a concern. As per this specific
>>> case, I'm going to fix it and push a V2.
>>
>> I understand that and in some cases it’s even narrower where one might just
>> be working on one distribution alone for single architecture for that case
>> failure on architectures they don’t care is
>> also a nuisance
>>
>> It’s wide and far testing we can achieve is good for end user same goes for
>> multiple architectures if someone tests a given architecture and reports
>> build  failures then it’s good to acknowledge that imo
>>
>> It’s not a blocker I expect someone interested in a given regression of
>> this kind to chime and help if possible so don’t see it as an impediment
> Is anyone actually using networkmanager with musl?
>
> There are no obvious usecases where it makes sense, and OE seems to be
> adding security vulnerabilities in the musl build that are not in the 
> glibc build (e.g. -Dsecure_getenv=getenv) so users should better not
> use it in any case.
>
>> If we remain cognizant of the fact that upstream might have wider usecases
>> we will generally be better off cumulatively
I have pushed a V2 which addresses the musl patches as well.
> Upstream ships a copy of systemd code for use as library.
>
> This already implies that trying to support it for musl will be a 
> long-term pain of not upstreamable patches for OE contributors.
That is worrying indeed.

-- 
Andrei Gherzan
gpg: rsa4096/D4D94F67AD0E9640 | t: @agherzan

Re: [meta-networking][PATCH] networkmanager: Update to 1.18.0

[Khem Raj]

On Tue, May 28, 2019 at 11:06 AM Adrian Bunk <bunk@stusta.de> wrote:
>
> On Tue, May 28, 2019 at 09:14:17AM -0700, Khem Raj wrote:
> > On Tue, May 28, 2019 at 8:37 AM Andrei Gherzan <andrei@gherzan.ro> wrote:
> >
> > > On 28/05/2019 16.13, Khem Raj wrote:
> > > > Hi Andrei
> > > >
> > > > The musl patches need to be forward ported as well. They dont apply
> > > > cleanly anymore on top of this update.
> > >
> > > Seems like. So all the contributions now need to go through local musl
> > > and glibc testing? This sounds a little bit of an overhead and might
> > > discourage people in contributing because usually everyone is working on
> > > one libc variant. Just raising it as a concern. As per this specific
> > > case, I'm going to fix it and push a V2.
> >
> >
> > I understand that and in some cases it’s even narrower where one might just
> > be working on one distribution alone for single architecture for that case
> > failure on architectures they don’t care is
> > also a nuisance
> >
> > It’s wide and far testing we can achieve is good for end user same goes for
> > multiple architectures if someone tests a given architecture and reports
> > build  failures then it’s good to acknowledge that imo
> >
> > It’s not a blocker I expect someone interested in a given regression of
> > this kind to chime and help if possible so don’t see it as an impediment
>
> Is anyone actually using networkmanager with musl?
>

Yes.

> There are no obvious usecases where it makes sense, and OE seems to be
> adding security vulnerabilities in the musl build that are not in the
> glibc build (e.g. -Dsecure_getenv=getenv) so users should better not
> use it in any case.
>

That could be improved. Probably adding this function to musl is not a bad idea

something like

char *secure_getenv(const char *name)
{
 return issetugid() ? NULL : getenv(name);
}

will do the job. Once I am back I might propose this to musl if someone does
that before would be fine too.

> > If we remain cognizant of the fact that upstream might have wider usecases
> > we will generally be better off cumulatively
>
> Upstream ships a copy of systemd code for use as library.
>
> This already implies that trying to support it for musl will be a
> long-term pain of not upstreamable patches for OE contributors.
>

so far it has not been so much a pain but its good to monitor.

> cu
> Adrian
>
> --
>
>        "Is there not promise of rain?" Ling Tan asked suddenly out
>         of the darkness. There had been need of rain for many days.
>        "Only a promise," Lao Er said.
>                                        Pearl S. Buck - Dragon Seed
>

Re: [meta-networking][PATCH] networkmanager: Update to 1.18.0

[Adrian Bunk]

On Tue, May 28, 2019 at 01:37:21PM -0700, Khem Raj wrote:
>...
> That could be improved. Probably adding this function to musl is not a bad idea
> 
> something like
> 
> char *secure_getenv(const char *name)
> {
>  return issetugid() ? NULL : getenv(name);
>...

I didn't know that musl adds non-standard functions like issetugid()
that are not in glibc.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: [meta-networking][PATCH] networkmanager: Update to 1.18.0

[Khem Raj]

On Tue, May 28, 2019 at 11:20 PM Adrian Bunk <bunk@stusta.de> wrote:
>
> On Tue, May 28, 2019 at 01:37:21PM -0700, Khem Raj wrote:
> >...
> > That could be improved. Probably adding this function to musl is not a bad idea
> >
> > something like
> >
> > char *secure_getenv(const char *name)
> > {
> >  return issetugid() ? NULL : getenv(name);
> >...
>
> I didn't know that musl adds non-standard functions like issetugid()
> that are not in glibc.

I don't think its stated as such anywhere in musl design guidelines,
however they do
try to implement defined behavior for POSIX functions which is a good thing.

>
> cu
> Adrian
>
> --
>
>        "Is there not promise of rain?" Ling Tan asked suddenly out
>         of the darkness. There had been need of rain for many days.
>        "Only a promise," Lao Er said.
>                                        Pearl S. Buck - Dragon Seed
>