Oops, I missed the “switching to desktop still works” part, indeed. Thanks.
Shouldn’t the end of the useragent be "); instead of >?
5 Likes
user_pref("general.useragent.override", "Mozilla/5.0 (X11; Linux aarch64; Mobile; rv:91.0) Gecko/20100101 Firefox/91.0");
Works with the correct ending as mentioned above.
1 Like
For me too. Weather is now working again
1 Like
Uh, you’re right, I fixed it in my post.
Thanks for your hint!
4 Likes
error: failed to determine package fingerprint for build script for selectors
I didn’t manage to fix the error but it was not introduced by a commit as navigating to commits I know that have built didn’t fix it, nor did cleaning.
I did manage to “nuke” the whole environment, by uninstalling the SDK and re-installing it. By using the online installer, I got the non-EA newer version 3.11.1 of the SDK.
This continued to challenge me with these 3 changes:
-
Specific to my setup - since I am running everything in an Ubuntu 20.04 VM (because the VM was set up to be a porting machine and that is what HADK uses for building Android bits), I was confronted with the fact that the SDK is now linked against
GLIBC_2_34—> I needed a “distribution upgrade” to Ubuntu 22.04 because of this. I didn’t find the specific symbol needed that version so I just assume a newer glibc was installed on the machine that builds the SDK. -
Since I am using 4.6 target images, I changed this spec line to read
libclang.so.15(see 4.6 Changelog) -
Finally, since the compiler was upgraded to 10.3 in the 4.6 OS/target (10.3.1 says the SDK link above) I needed to change this spec line to spell
/usr/include/c++/10.3.1/- the error thrown was inability to find “math.h” and the included path pointed toobj-build-mer-qt-xr/includewhich was linked to 8.3.0.
Let’s hope that I don’t have to search for changes 4. or 5. in the next hours and the build completes;)
(L.E. step 4 for fully debugging experience was to set chmod a+rx /home/.system where the symbols are → for some reason that was root-only access)
3 Likes
If memory serves this can be caused by building without the --with git_workaround flag (which you can add either directly or to your sfdk config).
I’m also going to have to restart with a fresh SDK to get the latest updates. Thank you for noting the changes you needed to make here as I can follow your advice now when I do.
3 Likes
As there are many here who installed the test packages and hit issues getting desktop versions of sites, I’ve uploaded fresh packages that include the user-agent changes needed to fix this.
Instructions for how to install them are in my latest post (very similar to the previous instructions).
Thanks to everyone who’s posted issues on GitHub. Please do continue to do this; it’s really useful.
16 Likes
Thanks for your good work, and as always clear instructions. 
1 Like
Thanks for the updated packages, sadly they broke the browser, on my 10 II it is crashing on startup deep in the wayland libraries according to the backtrace. Also apps using the webview are crashing.
Can anybody provide a copy of the old packages? In my wisdom I had overridden my files with the new ones.
1 Like
Sorry for this @balta and thanks for sharing the old packages @cypherpunks. I’ll look into it this evening and aim to post fresh tarballs.
6 Likes
I’ve updated the package which help the browser and webview apps to work correctly. As before the email app and settings app (account creation) are still likely to crash, but that shouldn’t be a change from the previous packages (fixing this requires an updated libybris).
8 Likes
Thanks for this update! Unfortunately browser won’t launch (I downloaded the packages 15 minutes ago, so they should be the updated ones, right?). E-Mail application does work, however.
The browser window appears, I see the address bar for a short moment and then it crashes and leaves a greyed-out cover window on the screen.
EDIT: Xperia 10 III, SFOS 4.6.0.13
1 Like
have you tried a reboot of the phone? had the same behavior, but after a reboot it worked.
1 Like
Yes I rebooted, but that didn’t help either.
1 Like
I can confirm that the new release does show the same behavior. Here is a not so helpful backtrace, I can try to create a more informative later at home:
Thread 42 "Compositor" received signal SIGSEGV, Segmentation fault.
[Switching to LWP 9836]
0x0000007faf16a814 in ?? () from /usr/lib64/libwayland-client.so.0
(gdb) bt
#0 0x0000007faf16a814 in ?? () from /usr/lib64/libwayland-client.so.0
#1 0x0000007faf16af54 in ?? () from /usr/lib64/libwayland-client.so.0
#2 0x0000007faf168908 in wl_proxy_marshal_flags () from /usr/lib64/libwayland-client.so.0
#3 0x0000007faee85660 in ServerWaylandBuffer::ServerWaylandBuffer(unsigned int, unsigned int, int, int, android_wlegl*, wl_event_queue*) ()
from /usr/lib64/libhybris//eglplatform_wayland.so
#4 0x0000007faee85700 in WaylandNativeWindow::addBuffer() () from /usr/lib64/libhybris//eglplatform_wayland.so
#5 0x0000007faee8444c in WaylandNativeWindow::dequeueBuffer(BaseNativeWindowBuffer**, int*) () from /usr/lib64/libhybris//eglplatform_wayland.so
#6 0x0000007faedb7240 in BaseNativeWindow::_dequeueBuffer(ANativeWindow*, ANativeWindowBuffer**, int*) ()
from /usr/lib64/libhybris-platformcommon.so.1
#7 0x0000007fac5908a0 in ?? ()
#8 0x0000007e841372c8 in ?? ()
Backtrace stopped: previous frame inner to this frame (corrupt stack?)
2 Likes
This problem looks like the hybris problem that was mentioned / commented at some point. As far as I recall, team pushed hybris fixes to the 4.6.0 branch meaning that next 4.6.0 update release will carry the fixes.
9 Likes
Thanks for all the helpful feedback from everyone on this. @rainemak is correct, the backtrace seems to relate to the hybris changes. There was an additional issue where a recent change broke the preference handling which I’ve now fixed. I’ve uploaded new packages which should now prevent the crash at start up, but which also have the user agent and TLS fixes integrated.
5 Likes