Segfault on navigator window launch

[ruskie]

Installed pale moon 26.2.1 x86_64 on Slackware -current(on a laptop). When starting it I get the select your profile window. Once I select a profile and launch it I get a segfault.
The same installation but with an older version of Slackware -current works without a problem(on a desktop).

Any futher info I can provide please ask.

Code: Select all

palemoon
segmentation fault  palemoon

Providing gdb output.

Code: Select all

(gdb) bt
#0  0x00007fffddb7a750 in  ()
#1  0x00007ffff5cd9f50 in JS_ObjectToOuterObject(JSContext*, JSObject*) ()
    at /opt/palemoon/libmozjs.so
#2  0x00007ffff325c3c7 in  () at /opt/palemoon/libxul.so
#3  0x00007ffff5c6cbc8 in  () at /opt/palemoon/libmozjs.so
#4  0x00007ffff5c6d648 in  () at /opt/palemoon/libmozjs.so
#5  0x00007ffff5c38bde in JS_WrapObject(JSContext*, JSObject**) ()
		    at /opt/palemoon/libmozjs.so
#6  0x00007ffff325cc7d in  () at /opt/palemoon/libxul.so
#7  0x00007ffff325cd83 in  () at /opt/palemoon/libxul.so
#8  0x00007ffff325ce57 in  () at /opt/palemoon/libxul.so
#9  0x00007ffff5d0948d in  () at /opt/palemoon/libmozjs.so
#10 0x00007ffff5d09649 in  () at /opt/palemoon/libmozjs.so
#11 0x00007ffff5b7abd5 in  () at /opt/palemoon/libmozjs.so
#12 0x00007ffff5b82d78 in  () at /opt/palemoon/libmozjs.so
#13 0x00007ffff5b77f7d in  () at /opt/palemoon/libmozjs.so
#14 0x00007ffff5b841f1 in  () at /opt/palemoon/libmozjs.so
#15 0x00007ffff5c3e9e4 in JS_CallFunctionValue(JSContext*, JSObject*, JS::Value, unsigned int, JS::Value*, JS::Value*) () at /opt/palemoon/libmozjs.so
#16 0x00007ffff323ac57 in  () at /opt/palemoon/libxul.so
#17 0x00007ffff37bc253 in  () at /opt/palemoon/libxul.so
#18 0x00007ffff37bb7d1 in  () at /opt/palemoon/libxul.so
#19 0x00007fffd3d4d040 in  ()
#20 0x00007fffd35f6080 in  ()
#21 0x00007fffd364aac0 in  ()
#22 0x00007fffd364aac0 in  ()
#23 0x00007fffffff9630 in  ()
#24 0x00007fffffff95e0 in  ()
#25 0x0000000000000000 in  ()

[Moonchild]

Did you build from source?
If I understood Ketmar correctly, Slackware is a little tricky to get Pale Moon to build correctly/in a stable fashion on.

[ruskie]

Nope. The slackbuild uses the binary tarball to produce a package:
http://linux.palemoon.org/files/26.2.1/ ... 64.tar.bz2

[Moonchild]

Exactly which versions of SlackWare are you talking about where it works/doesn't work?

[ruskie]

The one that is working is Slackware -current dated start of February, with no updates since.

The one that isn't working is Slackware -current dated April 24th.

Hope that helps.

[ketmar]

this is probably something i have to look into, but i don't have time and resources right now. i'm using 32-bit -current, though, so investigation can take even more time. sorry. i'll try to check it ASAP. stay tuned. ;-)

p.s. i used to do slackware packages for official slackware releases, built from sources (not repackaging). i'm planning to resume this activity too.

[ruskie]

Well if I can do anything to help out I can try. One thing I can do is try pale moon 32 bit in a 32 bit -current chroot. But I kinda suspect it'll work.

I thought that maybe one of the libraries got updated - the most obvious I've seen in my queue on the working system was glibc but I tried a downgrade with no luck. A few others as well but still the same issue.

[ruskie]

Well I've tried a few things - one was try the 32 bit build in the 32 bit chroot and that worked. The other was build it from source. It experienced the same segfault as the binary tarball. Tried a lower version of the binary tarball as well and got the same segfault.

[Walter Dnes]

When I build, my CFLAGS are set to

Code: Select all

"-O2 -march=native -mfpmath=sse -pipe -fomit-frame-pointer -fno-unwind-tables -fno-asynchronous-unwind-tables"

That's when I build on the machine it's going to run on. What are you using?

Another suggestion, requiring some external co-operation. I'm not running the standard 64-bit tarball, because I "roll-my-own". Can somebody who's successfully running 64-bit-linux standard build 26.2 copy+paste output from "about:buildconfig"? You'd need to compare that with the output on your machine, from...

Code: Select all

grep flags /proc/cpuinfo | head -1

Check if the build is invoking a register flag you don't have.

[gin83]

The segmentation fault is still there in palemoon-26.2.2_RC1.

I tried to build it with debug symbols, here is my ".mozconfig" :

Code: Select all

ac_add_options --build=x86_64-slackware-linux
ac_add_options --with-x
ac_add_options --enable-application=browser
ac_add_options --enable-official-branding
ac_add_options --enable-gstreamer=1.0
ac_add_options --disable-crashreporter
ac_add_options --disable-accessibility
ac_add_options --disable-parental-controls
ac_add_options --disable-gamepad
ac_add_options --disable-webrtc
ac_add_options --disable-logging
ac_add_options --disable-necko-wifi
ac_add_options --disable-installer
ac_add_options --disable-updater
ac_add_options --disable-websms-backend
ac_add_options --disable-windows-mobile-components
ac_add_options --disable-metro
ac_add_options --disable-maintenance-service
ac_add_options --disable-tests
ac_add_options --disable-mochitests
ac_add_options --enable-debug
ac_add_options --enable-debug-symbols
ac_add_options --disable-strip
ac_add_options --disable-jemalloc
ac_add_options --enable-optimize="-O2 -fPIC -ggdb"
ac_add_options --enable-safe-browsing
ac_add_options --enable-url-classifier
ac_add_options --enable-chrome-format=omni
ac_add_options --x-libraries=/usr/lib64

but these parameters don't work, I've got a build failure:

Code: Select all

Pale-Moon-26.2.2_RC1/caps/src/nsScriptSecurityManager.cpp:520:68: erreur: ‘const JSSecurityCallbacks {aka const struct JSSecurityCallbacks}’ has no member named ‘findObjectPrincipals’
         NS_ASSERTION(!JS_GetSecurityCallbacks(js::GetRuntime(cx))->findObjectPrincipals,

How can I build Palemoon with debug symbols, can you provide use with the parameters you use to build Palemoon with and without symbols?

Palemoon-26.2.2.RC1 builds with gstreamer-0.10 by default by the way.

Regards.

[trava90]

I can confirm debug builds fail on what appears to be the XSS filter. I've opened Issue #432.

Palemoon-26.2.2.RC1 builds with gstreamer-0.10 by default by the way.

I just checked, and can confirm that it is building GStreamer 1.x support (unless -enable-gstreamer=0.10 is used).

[Walter Dnes]

The one that is working is Slackware -current dated start of February, with no updates since.

The one that isn't working is Slackware -current dated April 24th.

Hope that helps.

Clutching at straws here. A couple of suggestions...
Plan A... rename the ".moonchild productions" directory, and start Pale Moon. See if that works.

Plan B... It looks like something has changed in Slackware. Trybuilding with as much Pale Moon internal stuff as possible. This would be in the hope that it avoidss whatever changed Slackware system-lib is causing the problem. I suggest jemalloc, jemalloc-lib, and "without-system-blah-blah-blah". Attached is a mozconfig file that works for me with 26.2_RelBranch. Note that you will have to make some changes.

1. can your system handle 'MOZ_MAKE_FLAGS="-j4"'
2. I have my build file automatically define an environmental variable ${OBJDIR} that MOZ_OBJDIR picks up. You'll have to set it manually
3. My build file also defines ${CFLAGS} and ${CCFLAGS}, which "--enable-optimize" picks up. You'll have to do it manually.

[ruskie]

So I've tried building with the provided mozconfig but using the normal howto build instructions. The result was sadly still a segfault after selecting the profile.

[dinosaur]

What is your gcc version ?

I just ran into issues with PM v26.2.2 complied (with gcc v4.9.2) for a newly installed PCLinuxOS 2016 system: segfaults every now and then when opening new tabs while on my previous distro (gcc v4.6.4) it ran rock-stable and never ever crashed.
I tried various options (without or with jemalloc, for example), but it kept crashing till I started incriminating the compiler itself... I tried to update to gcc-linaro v4.9.4, but to no avail. I then compiled gcc v4.6.4 for PCLinuxOS 2016 and compiled PM with it: no more crash...

If your compiler is newer than v4.8, you might also be interested in this page (i.e. try adding -fno-delete-null-pointer-checks to your compilation options).

Also, about gcc v4.9, have a look at what Linus Tovald himself said here.

[Walter Dnes]

Are you building with "-Os" (optimize for size)? If so, don't. According to post http://lkml.iu.edu//hypermail/linux/ker ... 00868.html

> 4.8 and 4.7 don't hit the problem on this test.
> 4.9 with -O2 compiles this file ok. 4.9 with -Os triggers it.

[dinosaur]

Are you building with "-Os" (optimize for size)? If so, don't. According to post http://lkml.iu.edu//hypermail/linux/ker ... 00868.html

> 4.8 and 4.7 don't hit the problem on this test.
> 4.9 with -O2 compiles this file ok. 4.9 with -Os triggers it.

I'm using -O3, but anyway, gcc v4.9 is now definitely a no-no (also hit the same kind of issue with openmotif compiled with gcc 4.9: random segfaults again... recompiled the libraries with gcc v4.6.4 and no more problem). I'll keep using v4.6.4 which works beautifully and never let me down in years, including on very large projects.

[ketmar]

-O3 is known to cause troubles too. stick with -O2, you won't notice much speed improvement from -O3 anyway. ;-)

[dinosaur]

-O3 is known to cause troubles too. stick with -O2, you won't notice much speed improvement from -O3 anyway.

-O3 works just fine, thank you. I have been using it for many years without issue whatsoever

[ketmar]

it worked fine with 4.6. then it causing unstable builds for 4.9. then i stop experimenting and stick with -O2. ;-)

[Walter Dnes]

I'm using -O3, but anyway, gcc v4.9 is now definitely a no-no (also hit the same kind of issue with openmotif compiled with gcc 4.9: random segfaults again... recompiled the libraries with gcc v4.6.4 and no more problem). I'll keep using v4.6.4 which works beautifully and never let me down in years, including on very large projects.

You may get away with it some of the time, but it's just asking for trouble. Even on the Gentoo support forums, home of us "Gentoo Ricers", -O3 is laughed at and unsupported. -O3 is famous for random segfaults, etc. I'm running gcc 4.9.3 and I've built Pale Moon successfully for my ancient Atom netbook, 2 desktops, and a notebook, using -O2. I repeat, do not use -O3.