purr data fails after pango upgrade

[Former user]

After upgrading pango to pango-1:1.44.3-1-x86_64, purr data fails to come up. Running with -nogui works.

The output of 'purr-data -verbose' is

Pd-L2Ork version 2.9.0 (20190507-rev.22573662) compiled 21:36:27 Aug 5 2019 port 5400 guidir is /opt/purr-data/lib/pd-l2ork/bin "/opt/purr-data/lib/pd-l2ork/bin"/nw/nw --user-data-dir="/home/morris/.config/purr-data" "/opt/purr-data/lib/pd-l2ork/bin" 5400 localhost pd-l2ork "/op priority 6 scheduling enabled. [0806/071308.573317:WARNING:chrome_main_delegate.cc(565)] final extension: Pd: signal 2

[Albert Graef]

Hmm, that doesn’t say much. I’m afraid that you’ll have to wait until I can install that pango version and try to reproduce that error.

Can you tell me the so version of the pango library that comes with pango-1:1.44.3-1-x86_64? On my system it’s libpango-1.0.so.0, and that’s also what nw.js links against.

Also, you could try running gdb purr-data and attach the backtrace here.

[Steve Morris]

I’m afraid this doesn't impart much more information. Current pango libs is

libpango-1.0.so.0.4400.3

All other pango frelated libs are at the same level.

gdb di not prove useful either.

$gdb purr-data

GNU gdb (GDB) 8.3

Reading symbols from purr-data...
(No debugging symbols found in purr-data)
(gdb) run

Starting program: /usr/bin/purr-data
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/usr/lib/libthread_db.so.1".
guidir is /opt/purr-data/lib/pd-l2ork/bin
[Detaching after fork from child process 26722]
priority 6 scheduling enabled.
[0807/082639.104403:WARNING:chrome_main_delegate.cc(565)] final extension:

C
Program received signal SIGINT, Interrupt.
0x00007ffff78d7b65 in accept () from /usr/lib/libpthread.so.0
(gdb) bt
#0 0x00007ffff78d7b65 in accept () from /usr/lib/libpthread.so.0
#1 0x000055555560019f in sys_startgui ()
#2 0x00005555555fd864 in sys_main ()
#3 0x00007ffff72f3ee3 in __libc_start_main () from /usr/lib/libc.so.6
#4 0x000055555556f03e in _start ()
(gdb)

[Albert Graef]

So it got stuck there until you typed Ctrl+C? That probably means that the nw.js GUI just refused to launch.

Can you please cd to /opt/purr-data/lib/pd-l2ork/bin/nw and run ./nw there? Does that display any error message?

[Steve Morris]

\$./nw
fish: “./nw” terminated by signal SIGSEGV (Address boundary error)

gdb further says

Thread 1 "nw" received signal SIGSEGV, Segmentation fault.
0x00007ffff40dd7ae in hb_face_reference_table ()
from /opt/purr-data/lib/pd-l2ork/bin/nw/lib/libnw.so

[Albert Graef]

Ah ok, we’re getting there. That routine there is part of harfbuzz which is being used by pango. So this might actually be a bug in the new pango version (which seems likely since it goes away if you simply downgrade pango). It could also be an nw.js issue which just happens to trip that segfault, but that seems less likely now.

Anyway, thanks a lot for the investigation. Now I actually have an idea about where to start looking, once the new pango version comes flying in (if it makes it through the Manjaro QA).

I’ll keep you updated. In the meantime, you will probably want to downgrade pango again to make purr work.

[Albert Graef]

I can reproduce the bug on Arch (EndeavourOS) with the old Purr Data package, but it goes away with the latest one, so it seems that indeed the bug is in nw.js 0.22.1, and updating the nw.js version to 0.24.4 solves the problem.

Can you confirm that updating to Purr Data 2.9.0 20190806-rev.c96563ce (which is now in the AUR and also in the binary package repo) fixes the problem?

[Steve Morris]

I tried it on my laptop and it appears to work, however my main machine has failed and I'm in the process of putting it back together. I'll follow up when I get my machine running again.

[Albert Graef]

Sounds good! Closing this bug for now, feel free to reopen it if needed.

Thanks for reporting!