Erratic Crashes in MX-Linux Vsn 15 [Solved]

[gordonC]

A few months ago, Pale Moon was added to the MX14 repository and has performed well for users, working with very different hardware. Now, MX-Linux is being upgraded, currently
at the second Beta testing stage. Pale Moon has been included in the MX15 repos.
If Pale Moon is installed from the MX15 repository, we are having crashes on some websites, while if PM is installed by downloading and installing from the Pale Moon source no crashes have occurred (yet). I acknowledge that an immediate answer could be that the problem is either in MX15 Beta, or in the repository package and that Pale Moon is okay.
The MX15 Pale Moon package has been upgraded to 25.8.1 with no improvement. As yet, the team has not found any obvious errors nor omissions in the package.
Crash behaviour is erratic. In some cases, connection to the desired site is not completed, in others connection is made but the crash occurs within a few seconds, some crashes
seem to initiated by mouse movement of a click, others happen without any user action. Sometimes, the next start of Pale Moon reports that Closedown was incorrect and presents
a Restore Screen.
We are still collecting data about the addresses involved, four are: Walmart, sbnation, Mozilla addons and Co-operative Bank, New Zealand. There are more yet to be identified.

My question is: has this odd crash behavior been noted with any other distributions, and/or in any previous developments of Pale Moon ? Suggestions will be very welcome.

[Moonchild]

The main question is: how was Pale Moon packaged/repackaged? Or was it built from source in the package -- and if so: how?

[New Tobin Paradigm]

Intermittent crashes... I have a bs theory (well only BS if it is wrong).

Do me a favor.. go into about:buildconfig and get me a screenshot.

[gordonC]

Have passed main query on to packager, awaiting reply. Please clarify screenshot required.

Gordon.

[New Tobin Paradigm]

Have passed main query on to packager, awaiting reply. Please clarify screenshot required.

Gordon.

Just a screenshot of about:buildconfig

[gordonC]

Sorry for my ignorance, dont't where to find about:config

[New Tobin Paradigm]

Sorry for my ignorance, dont't where to find about:config

Type about:buildconfig in the address bar in that build that has issues and take a screenshot of it and link or post it here.

[stevenpusser]

The buildconfig is in a patch that creates a mozconfig file, so here's what it reads:

Code: Select all

+export MOZILLA_OFFICIAL=1
+mk_add_options MOZ_CO_PROJECT=browser
+ac_add_options --enable-official-branding
+ac_add_options --enable-application=browser
+ac_add_options --disable-crashreporter
+ac_add_options --disable-accessibility
+ac_add_options --disable-parental-controls
+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-tests
+ac_add_options --disable-mochitests
+ac_add_options --disable-debug
+ac_add_options --disable-debug-symbols
+ac_add_options --enable-strip
+ac_add_options --enable-jemalloc
+ac_add_options --enable-optimize="-O3 -msse2 -mfpmath=sse"
+ac_add_options --enable-chrome-format=omni
+ac_add_options --x-libraries=/usr/lib
+ac_add_options --prefix=/usr

[New Tobin Paradigm]

Ok.. I have half of what I need.. But I want to see what version of gcc was used. So I need that from about:buildconfig

[gordonC]

about:buildconfig
Build Machine

Jessiex32
Build platform
target
i686-pc-linux-gnu
Build tools
Compiler Version Compiler flags
gcc gcc version 4.9.2 (Debian 4.9.2-10) -Wall -Wpointer-arith -Wdeclaration-after-statement -Werror=return-type -Wtype-limits -Wempty-body -Wsign-compare -Wno-unused -Wcast-align -std=gnu99 -fgnu89-inline -fno-strict-aliasing -ffunction-sections -fdata-sections -pthread -pipe -DNDEBUG -DTRIMMED -O3 -msse2 -mfpmath=sse -fomit-frame-pointer
c++ gcc version 4.9.2 (Debian 4.9.2-10) -Wall -Wpointer-arith -Woverloaded-virtual -Werror=return-type -Wtype-limits -Wempty-body -Wsign-compare -Wno-invalid-offsetof -Wcast-align -fno-exceptions -fno-strict-aliasing -fno-rtti -ffunction-sections -fdata-sections -fno-exceptions -std=gnu++0x -pthread -pipe -DNDEBUG -DTRIMMED -O3 -msse2 -mfpmath=sse -fomit-frame-pointer
Configure arguments

--enable-official-branding --enable-application=browser --disable-crashreporter --disable-accessibility --disable-parental-controls --disable-webrtc --disable-logging --disable-necko-wifi --disable-installer --disable-updater --disable-websms-backend --disable-tests --disable-mochitests --disable-debug --disable-debug-symbols --enable-strip --enable-jemalloc '--enable-optimize=-O3 -msse2 -mfpmath=sse' --enable-chrome-format=omni --x-libraries=/usr/lib --prefix=/usr

[New Tobin Paradigm]

Alright.. Firstly.. -O3 shouldn't be used.. kinda at all.. so change that to -O2. Secondly, --enable-jemalloc should NOT be used with GCC 4.9.

So.. Pass that along to the package maintainer.

BTW, pretty sure my theory is correct.

[gordonC]

Thanks Matt. Packager Mike will read this when he gets home from work.

[mmikeinsantarosa]

Hi Matt,
I'm the culprit that packaged up palemoon. Now about your advice;

Firstly.. -O3 shouldn't be used.. kinda at all.. so change that to -O2

I'm assuming this refers to

Code: Select all

--enable-optimize=-O3 

I do find this being set to 03 on line 1833 of configure.in

Code: Select all

MOZ_OPTIMIZE_FLAGS="-O3"

--enable-jemalloc should NOT be used with GCC 4.9.

I found enable-jemalloc in the configure.in of your source and it's mentioned 5 times, several of which are mixed in fairly well with other logic.

So just to be clear, these are the 2 changes you're talking about us making?

thanks
Mike

[trava90]

Hi Mike,

I'm assuming this refers to

Code: Select all

    --enable-optimize=-O3

Correct. There are known issues with random crashing when using -O3 and GCC 4.9+ on some systems.

I do find this being set to 03 on line 1833 of configure.in

Code: Select all

MOZ_OPTIMIZE_FLAGS="-O3"

That specific line is for Mac OS.

I found enable-jemalloc in the configure.in of your source and it's mentioned 5 times, several of which are mixed in fairly well with other logic.

enable-jemalloc will no longer be used beginning with Pale Moon 26. On modern Linux systems, the OS memory allocator is as good or better than jemalloc (i.e. you will get better performance if you don't use jemalloc).

So just to be clear, these are the 2 changes you're talking about us making?

Yes, these should solve the problem.

(Sorry for jumping in Matt)

[stevenpusser]

Mike, those settings are in the mozconfig.patch file. You can safely edit the -03 to 02, but if you delete any lines, you'll have to adjust that final line count of 23 in the header section of that patch.

[ketmar]

jemalloc, as far as i tested on various gcc versions and systems, is not a culprit (althru it serve almost no purpose, as glibc allocator is good enough these days). but -O3... yes, this is what causes random crashes at gcc 4.9+.

so far, with jemalloc turned off and -O2, i found no problems (and no crashes) on gcc 4.9.2, gcc 4.9.3, and gcc 5.2. usualy my Pale Moon sessions lasts for weeks (i'm not turning off computer and not hibernating it), without any crash at all.

[gordonC]

Fixed we think. Changes proposed by Matt have been made and crashes at this location have stopped. Thanks to all involved.

[mmikeinsantarosa]

Thanks to all involved.

+1 - mike