Sumatra (portable version 3.0) crash about sumatrapdf

[kjk]

Where did you download Sumatra from?

I couldn't reproduce this crash with http://www.stiltz.co.uk/assets/downloads/duo_footprint_tech.pdf in 3.0 (neither the installer nor portable).

It could have been caused by lack of memory.

We should already have the crash report, I just don't know which one would be from you. If you can tell me the ip address of your computer (visit http://www.myipaddress.com/show-my-ip-address/), then I can try to find the crash report that came from that machine.

[JNicoll]

You wrote:

Where did you download Sumatra from?

http://www.sumatrapdfreader.org/download-free-pdf-viewer.html

• the 'portable' link. I renamed the file after downloading it
  and it now has the following characteristics:

Filename : 20141211 1626 SumatraPDF_3-0.zip
File Size : 3,312,542
MD5 : 00a08d45de5154894615c2c040f232ad
SHA1 : 61b61d01bfa7a8e8e34484b9e2a2e6fc48647feb
CRC32 : a610cfdc
SHA-256 :
083f710589f48792f65563764e964ed7bcda947037100d9b2c12fc1519e222d6
SHA-512 :
87903bc345f17ee6399622cc1c79fe2ad53a502788b4600faaf967f287dc14c118baf030f375f08985fc2c35fbdbe4c636b29741fb0201099740720151526636
SHA-384 :
d62c2ea967cc7117302e0ff385e26a79c4f92ee070fe8504dcd9c7413ca355bb1d149f61c8e6b0a32efbca5f12956471
Full Path : C:\My Dropbox\Downloads\2014\20141211 1626
SumatraPDF_3-0.zip

I couldn't reproduce this crash with
http://www.stiltz.co.uk/assets/downloads/duo_footprint_tech.pdf in 3.0
(neither the installer nor portable).

It could have been caused by lack of memory.

Shouldn't Sumatra handle that rather better? ;-)

The laptop's got 2 GB RAM, which is reasonably generous for an XP box. I
can't remember what else I was doing at the time though.

We should already have the crash report, I just don't know which one would
be from you. If you can tell me the ip address of your computer:

82.42.156.97 at the moment, hopefully the same then. The machine's been
rebooted since, but the internet connection itself hasn't been off.

The crash happened at approx 20141222 17:56 (UK time, ie about 9 hours ago).

Thanks for such a quick reply!

---

Reply to this email directly or view it on GitHub:
#53 (comment)

Jeremy Nicoll - my opinions are my own.

[kjk]

Can you tell us how do you load the document?

The crash seems to be:

Ver: 3.0
Browser plugin: no
OS: Windows XP SP3 build 2600 32-bit
Number Of Processors: 2
Processor:          Intel(R) Atom(TM) CPU N270   @ 1.60GHz
Physical Memory: 1.99 GB
Commit Charge Limit: 4.83 GB
Memory Used: 63%
Lang: en GB
Graphics driver 0
  DriverDesc:         M
  DriverVersion:      6.14.10.4926
Graphics driver 1
  DriverDesc:         M
  DriverVersion:      6.14.10.4926

Exception: C0000005 ACCESS_VIOLATION
Faulting IP: 004336EC 01:000326EC sumatrapdf.exe!UpdateCurrentTabBgColForWindow+0x4c c:\users\kkowalczyk\src\sumatrapdf30\src\tabs.cpp+719
Fault writing address 00000088

Registers:
EAX:00000000  EBX:00000000  ECX:7E4188DA
EDX:0043D001  ESI:0179A8A8  EDI:FFFFFFFF
CS:EIP:001B:004336EC
SS:ESP:0023:0012F874  EBP:00000000
DS:0023  ES:0023  FS:003B  GS:0000
Flags:00010246

Crashed thread:
004336EC 01:000326EC sumatrapdf.exe!UpdateCurrentTabBgColForWindow+0x4c c:\users\kkowalczyk\src\sumatrapdf30\src\tabs.cpp+719
0043396D 01:0003296D sumatrapdf.exe!TabsOnLoadedDoc+0x9d c:\users\kkowalczyk\src\sumatrapdf30\src\tabs.cpp+797
0040588C 01:0000488C sumatrapdf.exe!LoadDocumentOld+0x32c c:\users\kkowalczyk\src\sumatrapdf30\src\sumatrapdf.cpp+1560
00411229 01:00010229 sumatrapdf.exe!LoadOnStartup+0x69 c:\users\kkowalczyk\src\sumatrapdf30\src\sumatrastartup.cpp+205
004121A4 01:000111A4 sumatrapdf.exe!WinMain+0x4c4 c:\users\kkowalczyk\src\sumatrapdf30\src\sumatrastartup.cpp+527
0065E96B 01:0025D96B sumatrapdf.exe!__tmainCRTStartup+0x11a f:\dd\vctools\crt_bld\self_x86\crt\src\crt0.c+275
7C816037 01:00015037 kernel32.dll!RegisterWaitForInputIdle+0x49

It doesn't seem to be related to running out of memory or this specific document but to how the document is loaded.

[JNicoll]

You wrote:

Can you tell us how do you load the document?

I just double-clicked the file in Windows Explorer.

Exception: C0000005 ACCESS_VIOLATION
Faulting IP: 004336EC 01:000326EC
sumatrapdf.exe!UpdateCurrentTabBgColForWindow+0x4c
c:\users\kkowalczyk\src\sumatrapdf30\src\tabs.cpp+719

Does this 'tabs.cpp' mean the fault's in the new(?) tabs handler? I've got
my Sumatra configured to use the old-style not-tabbed interface.

Jeremy Nicoll - my opinions are my own.

[kjk]

Even when we don't show tabs this code is shown.

I've tried to reproduce this without tabs in 3.0 and it didn't crash for me. It's perplexing.

Does it always happen for you with this document?

Can you try pre-release version http://www.sumatrapdfreader.org/prerelease.html. We've made changes in that logic.

[zeniko]

The crash seems to be due to GetWindowLongPtr(win->hwndTabBar, GWLP_USERDATA) returning NULL in SetCurrentTabBgCol. Could it be that either Windows or a third-party tool overwrites the value we set in CreateTabbar?

[kjk]

I guess that's possible but wouldn't that cause crash for all documents?

I think only in-process software could do that. At some point I moved enumeration of dlls reported in crash reports from time of crash to startup, which might no longer report dlls injected at a later point. I should re-add this information to a crash report.

[JNicoll]

You wrote:

Even when we don't show tabs this code is shown.

I've tried to reproduce this without tabs in 3.0 and it didn't crash for
me. It's perplexing.

Does it always happen for you with this document?

It happened twice at least yesterday, with Sumatra 3.0; after that I didn't
try again. If it's a lack of memory issue maybe something else was wrong
with XP at that point and Sumatra was a victim rather than the cause.

Having said that in the tests described below, task manager showed Sumatra's
MemSize of only 14 - 15.5 MB for each version when the test file was being
displayed, and VM sizes around 7-8 MB. It's hard to see that as a threat!

Just after the crash yesterday explorer stopped working - in the sense that
scripts I have which I use all the time to open regularly-used folders
(which issue explorer.exe commands) no longer opened new folder views,
though existing explorer windows stayed open. I shut the machine down soon
after that.

I've looked at the event logs and there's nothing either from Sumatra at the
time of the crash, or anything else in the minutes beforehand to suggest a
system problem.

I also just looked at the logs from my anti-malware program (Emsisoft's
'EAM') and it shows nothing being inhibited, and no rules update happening
around that time - that's been known to stiff the machine with, it 'feels'
like, something then preventing new processes from being created.

I also just looked at my firewall / HIPS app - Emsisoft's OA (Online Armor);
it shows the same 'Allow', 'Trust' for v3.0 as it does for the three others
I've just tested.

I downloaded the drawings file from the Stilz website again (because I'd
deleted it after the crash), and tried to open it in portable versions 2.2.1
and 2.3.2 of Sumatra, those being prior versions I've used in the past which
I still had the downloads for. In both cases the app was started in its own
temporary folder so did not pick up the settings file that my v3.0 one uses.
In both cases I started the app, used File->Open to open the PDF, and it
displayed fine.

I then tried the pre-release 10082 app, running out of a third folder. For
this I started the app, changed its settings to turn off tabs, closed the
app, restarted it, used File->Open ... and this time the file displayed
properly and immediately.

Finally I tried v3.0 again... and it also worked.

I noted someone's comment about injected DLLs. I presume EAM & OA will be
responsible for some; I also typically have 'Process Hacker' running, to
give me a better idea (than TM does) of what's going on in the machine. I
assume that injects some DLL into every process too. But it was also
running when I did today's tests...

I'm not knowledgeable on Windows internals (but used to be an MVS - now
what's known as z/OS - systems programmer and also worked as a programmer on
IBM's MVS and VM systems). If there's any diagnostic process you want me to
do, I'm happy to try. I do see that if Sumatra now wants to behave
perfectly that's not going to be helpful.

Reply to this email directly or view it on GitHub:
#53 (comment)

Jeremy Nicoll - my opinions are my own.

[kjk]

Not enough information to make progress.