On 12/17/25 05:12, Ross Burton wrote:
CAUTION: This email comes from a non Wind River email account!
Do not click links or open attachments unless you recognize the sender and know
the content is safe.
On 16 Dec 2025, at 10:49, hongxu via lists.openembedded.org
<[email protected]> wrote:
On 12/12/25 21:39, hongxu via lists.openembedded.org wrote:
While installing 32 bit gstreamer package and 64 bit gstreamer package to
one rootfs, package management (dnf) reported a conflict failure
|Error: Transaction test error:
| file /usr/share/gir-1.0/Gst-1.0.gir conflicts between attempted installs of
gstreamer1.0-dev-1.26.7-r0.wr2500.corei7_64 and
lib32-gstreamer1.0-dev-1.26.7-r0.wr2500.corei7_32
Due to upstream commit [1], the size of padding differs between 32 bit
arch and 64 bit arch
The commit doesn’t exist anymore so that URL is 404,
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/commit/0ee18781f9b210129da23a85236266bf84bafa83
is a working link.
OK, I will update the link
This is a bit of a complex problem and just mv'ing the file isn’t all of it.
It would be easier to just mv it in a do_install:append() considering upstream
doesn’t seem interested. But the bigger problem is how does this interact with
the rest of the G-I ecosystem? Is our G-I configured to search that location?
What about PyGObject?
This patch reference glib [1][2][3], the meson option gir_dir_prefix:
'Intermediate prefix for gir installation under ${prefix}'
[1]
https://gitlab.gnome.org/GNOME/glib/-/blob/main/girepository/meson.build?ref_type=heads#L24
and correctly install .gir files into custom locations [2]
[2]
https://gitlab.gnome.org/GNOME/glib/-/commit/22ec5a96e3d7b1e17a4b0c31e00089ed17a12050
Absolutely worth referencing
https://gitlab.gnome.org/GNOME/gobject-introspection/-/issues/323 where the
core issue in G-I is fixed, by adding support for GI_GIR_PATH. So if we’re
going to move .gir files around, that needs to be updated too.
and compile with option -DGIR_DIR= [3] for C
[3]
https://gitlab.gnome.org/GNOME/glib/-/blob/main/girepository/meson.build?ref_type=heads#L88
Also, does this _actually_ work? Do we have on-target multi-arch
g-ir-compilers so you can have a 64-bit rootfs with a 32-bit g-ir-compiler that
does the right thing? Bandaiding a solution that breaks other users, and
doesn’t actually do anything useful, isn’t a solution.
I did not test it actually, yes, you are right, I will send v3 if
runtime test passed, otherwise just use original v1 [ gstreamer1.0: add
mulitlib scripts Gst-1.0.gir], treat as a multilib script
Alex, what is your option?
//Hongxu
Ross
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#227994):
https://lists.openembedded.org/g/openembedded-core/message/227994
Mute This Topic: https://lists.openembedded.org/mt/116745751/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
