[meta-swupdate][PATCH] Add libgcc to list of runtime dependencies

[Alexander Vickberg]

libgcc_s.so is dlopened by glibc so bitbake cannot automatically add

it. Without this library on a minimal image swupdate outputs the

following: "libgcc_s.so.1 must be installed for pthread_cancel to

work". Until a method to detect usage of functions in glibc which

dlopen libgcc_s.so is implemented in bitbake it is necessary to add it

manually with RDEPENDS.

Please see this discussion for more information:

https://yocto.yoctoproject.narkive.com/Hi849dZ8/btrfs-tools-requires-libgcc-s-so-1

Signed-off-by: Alexander Vickberg <wickbe...@gmail.com>

---

 recipes-support/swupdate/swupdate.inc | 1 +

 1 file changed, 1 insertion(+)

diff --git a/recipes-support/swupdate/swupdate.inc b/recipes-support/swupdate/swupdate.inc

index 700d5a1..6be7313 100644

--- a/recipes-support/swupdate/swupdate.inc

+++ b/recipes-support/swupdate/swupdate.inc

@@ -2,6 +2,7 @@ SUMMARY="Image updater for Yocto projects"

 DESCRIPTION = "Application for automatic software update from USB Pen"

 SECTION="swupdate"

 DEPENDS = "libconfig"

+RDEPENDS_${PN} = "libgcc"

 LICENSE = "GPLv2+"

 LIC_FILES_CHKSUM = "file://COPYING;md5=0636e73ff0215e8d672dc4c32c317bb3"

 

-- 

2.17.1

[Jeroen Hofstee]

Hi,

On 12/5/19 8:33 AM, Alexander Vickberg wrote:
> libgcc_s.so is dlopened by glibc so bitbake cannot automatically add
> it. Without this library on a minimal image swupdate outputs the
> following: "libgcc_s.so.1 must be installed for pthread_cancel to
> work". Until a method to detect usage of functions in glibc which
> dlopen libgcc_s.so is implemented in bitbake it is necessary to add it
> manually with RDEPENDS.
>

If glibc dlopens this, shouldn't it be a RDEPEND of glibc then?

Regards,

Jeroen

[Alexander Vickberg]

Hi Jeroen,

This was mentioned in https://bugzilla.yoctoproject.org/show_bug.cgi?id=10954. However Glibc only dlopen this in certain functions so it seems to not be done to avoid bloating the image. It also says that the recipes shouldn't have to do this but for now it seems to be the only solution. The link in the commit message is about btrfs-tools which faced the same issue and there it was solved by adding libgcc to RDEPENDS.

Best regards

Alexander Vickberg

[Jeroen Hofstee]

Hello Alexander,

On 12/5/19 10:03 AM, Alexander Vickberg wrote:
>
>
> This was mentioned in
> https://bugzilla.yoctoproject.org/show_bug.cgi?id=10954. However Glibc
> only dlopen this in certain functions so it seems to not be done to
> avoid bloating the image. It also says that the recipes shouldn't have
> to do this but for now it seems to be the only solution. The link in
> the commit message is about btrfs-tools which faced the same issue and
> there it was solved by adding libgcc to RDEPENDS.
>

Then they should just make it optional and default to depend on it.
Those who want to squeeze the last kilobytes out of an image
can then disable it, but at least you get a normal functional library
by default.

Regards,

Jeroen

[Stefano Babic]

I agree with Jeroen, and it is weird that an issue (or a missing
dependency) in another package should be fixed here. However, I see this
as workaround not just in btrfs, but in many other ones (ntp, tpm2, ..).
So I guess for now it is ok to merge this, but still open for a cleaner
solution in future, if any.

Regards,
Stefano



--
=====================================================================
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-53 Fax: +49-8142-66989-80 Email: sba...@denx.de
=====================================================================

[Ayoub Zaki]

Hi,

be aware that there are people who do not use gcc but clang specially to avoid having libgcc in their image!

BR

[Stefano Babic]

Hi Ayoub,

On 10/12/19 16:18, 'Ayoub Zaki' via swupdate wrote:
> Hi,
>
> be aware that there are people who do not use gcc but clang specially to
> avoid having libgcc in their image!

Thanks for pointing out - that is correct. There is also the use case
without glibc at all - SWUpdate is sued at least with musl. I am also
unsure if adding libgcc is ok in case of musl (or in case of another
glibc flavour). Anyway, I do not understand how this is managed by other
packages. Taking into account btrfs (well, not very often on embedded, I
admit) or better ntp: libgcc is then added to RDEPENDS. I would prefer
to move it into the image's recipe (rootfs), and anyone can decide if it
is required or not. Maybe adding some hints to the documentation.

Regards,
Stefano

>
> BR
>
>
>
> On Tuesday, December 10, 2019 at 12:13:04 PM UTC+1, Stefano Babic wrote:
>
> On 05/12/19 10:03, Alexander Vickberg wrote:
> > Den torsdag 5 december 2019 kl. 09:11:00 UTC+1 skrev Jeroen Hofstee:
> >
> >     Hi,
> >
> >     On 12/5/19 8:33 AM, Alexander Vickberg wrote:
> >     > libgcc_s.so is dlopened by glibc so bitbake cannot
> automatically add
> >     > it. Without this library on a minimal image swupdate outputs
> the
> >     > following: "libgcc_s.so.1 must be installed for
> pthread_cancel to
> >     > work". Until a method to detect usage of functions in glibc
> which
> >     > dlopen libgcc_s.so is implemented in bitbake it is necessary to
> >     add it
> >     > manually with RDEPENDS.
> >     >
> >
> >     If glibc dlopens this, shouldn't it be a RDEPEND of glibc then?
> >
> >     Regards,
> >
> >     Jeroen
> >
> >
> > Hi Jeroen,
> >
> > This was mentioned
> > in https://bugzilla.yoctoproject.org/show_bug.cgi?id=10954

> <https://bugzilla.yoctoproject.org/show_bug.cgi?id=10954>. However

> > Glibc only dlopen this in certain functions so it seems to not be
> done
> > to avoid bloating the image. It also says that the recipes shouldn't
> > have to do this but for now it seems to be the only solution. The
> link
> > in the commit message is about btrfs-tools which faced the same issue
> > and there it was solved by adding libgcc to RDEPENDS.
> >
>
> I agree with Jeroen, and it is weird that an issue (or a missing
> dependency) in another package should be fixed here. However, I see
> this
> as workaround not just in btrfs, but in many other ones (ntp, tpm2,
> ..).
> So I guess for now it is ok to merge this, but still open for a cleaner
> solution in future, if any.
>
> Regards,
> Stefano
>
>
>
> --
> =====================================================================
> DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
> HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
> Phone: +49-8142-66989-53 Fax: +49-8142-66989-80 Email:

> sba...@denx.de <javascript:>
> =====================================================================
>
> --
> You received this message because you are subscribed to the Google
> Groups "swupdate" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to swupdate+u...@googlegroups.com
> <mailto:swupdate+u...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/swupdate/08238887-d02e-477d-bf32-cef50272edc5%40googlegroups.com
> <https://groups.google.com/d/msgid/swupdate/08238887-d02e-477d-bf32-cef50272edc5%40googlegroups.com?utm_medium=email&utm_source=footer>.