This is followup from GEMESYS... Firefox on Windows fails regularly now, several times per day. Moving to different platform (CentOS (Fedora Linux), and new dev. work now done on Apple Mac, running Yosemite). Note: Firefox on both those new platforms seems to be stable. CentOS is a dreamland box... built it as an experiment, and the thing runs so well, is so stable, so nice and easy to use, (no weirdly awful GUI grief, like on the Mac..), that it has been rolled into full production. Runs 24/7. Problem free. No FF crashes, none. The Linux box just does not crash. It's an old 32-bit Pentium, stable, fast and easy - like my old 20th century Camaro -ie. turn key, go drive...).

Using Firefox on Windows now is like being on a Germanwings flight from Spain, with young Herr Lubitz at the controls. You know what will happen, it is just a question of when. We *really* need to fix this... The problem is *not* Windows. The problem is Firefox 41.0.

Specifically: The error now generated in reports is:

"[D3D9] Could not Initialize the DeviceManagerD3D9"

Searching the code, reveals this error message generated by codeframe:

 " gfxWindowsPlatform.cpp"

on line: 1532.

This c++ program element is 2880 lines long. The code may be accessed via URL: https://dxr.mozilla.org/mozilla-central/souce/gfx/thebe/gfxWindowsPlatform.cpp?offset=600

I have updated all drivers, and my HP WindowsXP/SP3 box runs a wide range of other graphics programs correctly, as I have indicated.

Again, if I can track down what is causing the crashes, I will post a fix, or a workaround here. Please, if anyone has encountered this issue with program module "gfxWindowsPlatform.cpp", send me a pointer to the corrective details. Thx.

In the address bar, type about:crashes<enter>. Note: If any reports do not have BP in front of the numbers/letters, click it and select Submit.

The crash report is several pages of data. We need the report numbers to see the whole report.

Using your mouse, mark the most resent 7 - 10 crash reports, and copy them. Now go to the reply box below and paste them in.

FredMcD, thank you for this information. There are about 100 crash reports *without* the "bp-" in front of the hexadecimal string. I was only able to submit the first one (Firefox 41.0.1 crashed again this morning, when I started it.) But I am using it to type this. Crashing is random, and does not seem related to any action on my part, other than web-browsing and reading. Specifically, the crash ID I was able to submit (I think), is URL as follows:

https://crash-stats.mozilla.com/report/index/453ea39b-a687-427b-a1fe-ff7e62151006

I note the crash reason seems to be "read-access" error of some kind. Is this assumption correct? I have disabled hardware-acceleration, to see if that might reduce crashing. It did not seem to make a difference. I also removed Firefox completely (as much as is possible), reload old 3.0.14, then upgraded to 3.0.19, and then from that version, downloaded the install stub for 41.0.1, and re-loaded full 41.0.1 version of Firefox for WindowsXP. I have similar Firefox 41.0.1 with current Flash plugin, installed on an older DELL laptop, as a video-viewer setup. We can watch any video on offer, including live streaming from many sites (eg. NHK World English from Japan, etc..), and that configuration *never* crashes. It is super stable, so FIrefox 41 failures on my HP Desktop machine (dual Pentium processor), is not purely a WindowsXP issue, as the little DELL laptop is rock solid crash free.

I have Microsoft .NET framework stuff installed, a couple of different version, installed recently as part of what I recall was a Skype upgrade. There is also a Java JRE runtime version (7.x or something) installed on this machine, as well as a couple of development environments (Blackberry (early Playbook stuff) and Android 4.x stuff).

Any suggestions, please let me know. Thx. (Yesterday, 4 or 5 crashes. Provoked the complete re-install of Firefox 41, and I verified the various .exe installers, checking for errors. All looks ok, etc. )

Ok, second crash now has "BP-" in front of it, so I am assuming it has been submitted. Here are top few crash hex strings...

bp-c27c5028-e0a0-438a-9377-44e632151006 10/6/2015 12:30 PM bp-453ea39b-a687-427b-a1fe-ff7e62151006 10/6/2015 12:30 PM c3062074-dfef-4745-827f-478bbe9e35f8 10/5/2015 11:00 PM e8a264d1-f395-41ea-8ce3-3ecaf8ea613f 10/5/2015 5:00 PM 1baec3d3-e338-4914-b1ab-21b07314eac7 10/5/2015 4:57 PM ef0ca798-9e9a-489e-93bc-d1e7ab946a44 10/5/2015 3:37 PM 23add49d-03d0-467c-aa6f-f1a13e5fe526 10/5/2015 3:01 PM dbe4a944-c447-4061-9404-e90a9fd40ab3 10/5/2015 2:09 PM 430578c6-8a85-405c-b3c1-aee43be56faa 10/4/2015 11:14 PM dfb49676-41d0-4c98-b7ca-fa11c2feac20 10/1/2015 10:34 PM 299794a6-4f11-482e-bcd8-0a170e2e9d6f 10/1/2015 3:58 PM d480c20f-724b-4780-a363-52c50e789530 10/1/2015 1:28 PM 8f492ea3-32c3-4f0b-b246-47536ffcfb33 9/30/2015 1:38 PM b676971e-5efd-4995-b6c3-1d31cb68b137 9/30/2015 10:28 AM

Hope this is helpful. Whenever crash occurs, I *always* get "There was a problem submitting your report", so no crash telemetry is being sent to Mozilla. I have the boxes checked to do the crash report submission, in Firefox Options.

FYI: Have submitted bugzilla bug, ID: 1211990

1211990 Firefox Untriaged nobody@mozilla.org UNCO --- Repeated, ongoing random crashes, Firefox 41.0.1, not plugin related. Two crash reports submitted. 10:34:50

Looking for a workaround for now. At very least suggest crash details (ie. exactly what failed) be provide to user, so they can attempt workaround. For example, I disabled "Hardware Acceleration:" This has not helped, but I have ruled out the "Driver Mismatch" message being generated when "Hardware acceleration" was operational.

I think we can mark this *SOLVED*. I downloaded the .iso disk image of "Memtest86+-5.01.exe", and used my new MacBook to burn a disk image of "Memtest86.exe", and confirmed (on a second box) that the MemTest86 program was testing memory correctly. This whole random crashing thing was disturbing. I was blaming Firefox, because that is what I ran on this machine, 90% of the time.

Turns out it looks to be a memory error. MEMTEST86 which runs to completion on other machines, produces big red error messaage at 1.1 gig (of a 2 gig RAM address space). Looks like that nails the problem. First time in 20 years I have had memory go bad like this. An image of the MEMTEST86 screen display is provided:

Note: Here is how to do memory testing..

   1) Get MEMTEST86+-5.01.zip or .tar.gz from one of the MEMTEST86
       sites. (Google or DDG it, there are lots...)

   2) Extract the MEMTEST86+_5.01.iso  file and put it somewhere.
       Files called: *.iso are typically disk images.  It's less than three
       meg.

   3) The trick with memory testing is to run this utility on the bare 
       machine, without any operating system.  That means you need
       to burn this MEMTEST86...iso file to a CD, and then boot your 
       machine from that CD.  So, find some way to burn *directly* 
       this *.iso file to a CD.   I used a new Apple MacBook I just got.
       I ruined a first disk, by burning the *folder* with the MEMTEST86
       program file in it to disk.  Of course, that disk is useless, as it
       is not bootable.  You need to run, (if using a MacBook), the Disk
       Utility program.  You can find it by clicking on the goofy "Finder"
       program, (in the "dock"), or on the top bar, clicking "Finder
       Preferences" and make sure "Hard DIsks" is checkmarked.
       (This will let you "Find" against the "Macintosh HD".)   When you
       run Finder on the whole HD, you will see folder "Applications".  
       Click, and then click "Utilities".  In that folder, you will find the 
       "Disk Utility.app"  That is the one you want.
 
        CLick on it, and it will give you a "Burn" option that has a yellow
       and black icon at the menu panel top that looks like a radioactive
        waste symbol.  That is the one to use.

       Click "Burn", and select your file MEMTEST86+_5.01.iso, and burn
       that file to a blank CD.  This will make a bootable CD which runs 
       nothing but MEMTEST86.  Just boot it, and it will start running.

   4) Let it run overnight if you want.  There should be *NO* red error
       messages.   In my case, with my HP-WindowsXP/SP3 box, the red
       "MEMORY FAIL" messages started coming up almost immedately,
       and were coming with *every* pass, consistantly.

It looks like things would run well, until something hit the 1.15 gig level. I have turned on a hardware feature to prevent data execution, but I don't think this would account for the errors.

Image of the MEMTEST86.exe results provided. Hope this helps others with random crash problems. Check your hardware.

That was very good work. Well Done. Please flag your last post as Solved Problem so others will know.

Go back to your bug report, and give details of how you fixed it.

I wanted to run a few checks, and turn off the hardware data execution protection (as per original machine config) before I marked this as solved. I also wanted to run the memory tests again, and confirm no errors. Have done this, and looks like this crashing issue is solved, and was due to bad memory.

The machine was 1.5 gig RAM (as one can obviously see from above MEMTEST. Removed the 512 meg RAM bank, and machine passes all the MEMTEST86 checks. Attached are three pics, the HP box board with just the 1 gig bank of the good, the bad 512 meg bank, now removed, and the resulting "Blue Screen of Happiness", courtesy of MEMTEST86.

Currently running Firefox 41.0.1 on WinXP/SP3, with 8 tabs open. Have run multiple tests, and could not provoke a crash. Marking it *solved*.

As Tony the Tiger says; It's Grrrrreat! !

Remember the bug report.