[Helixhorned]

Forge, on 28 March 2013 - 06:44 AM, said:

extreme circumstances, but it's reproducible.
walking through vent systems is where this is normally encountered

walk out on top of a floor aligned vent grate sprite/fan sprite, crouch and look all the way down, shoot or kick the sprite to break it, game crashes when duke starts to fall through the sprite

I can't reproduce this, but r3555 is also a bit oldish. Does it still happen with r3627?

Quote

edit: kinda funny that .crashlog and .map extensions aren't authorized attachments

It will be named ...crash.log now.

[Forge]

updated to 3627

it still does it, but the frequency seems more random & the way it happens changed
you no longer get all the way out onto the sprite
you basically have to step onto it right at the edge to where if you shoot it you won't fall through
the while crouched and looking as far down as possible, step forward

i could no longer make it crash with the fan sprite, just the vent sprite.

[Paul B]

Just playing Eduke on my wife's laptop. I was about 20 minutes into E1L1 and the game crashed to the desktop. The version of Eduke is: eduke32_win32_20130328-3621

Please see the attached crash log maybe this might help sort out the bug. No custom Mods or con files are being used or any other third party addon.

Attached File(s)

•  eduke32.log (5.75K)
  Number of downloads: 587

[TerminX]

Whoa, that's a nasty one. Unfortunately, there's nothing in that log that does anything to indicate the problem. Can you play for a while again, this time using the debug executable? Try to do the same things you did previously, particularly whatever you were doing as it crashed.

[Paul B]

TerminX, on 30 March 2013 - 10:41 AM, said:

Whoa, that's a nasty one. Unfortunately, there's nothing in that log that does anything to indicate the problem. Can you play for a while again, this time using the debug executable? Try to do the same things you did previously, particularly whatever you were doing as it crashed.

Yea the crash removed all the keyboard bindings which I had to reset to their defaults. I played the map again and was successfully able to finish the level. However, after playing E1L2 for about 20 minutes it froze up again and dumped me back to the desktop. I was running the debug version of eduke this time. Please see the attached logs.

I will make a comment about the laptop. I find these crashes could be a result of the laptop getting too warm. This may be a result of hardware as oppose to Eduke itself as Eduke really heats up the laptop when playing which as a result seems to make the keyboard/touchpad to hesitate when pressing a key or moving the cursor. I wouldn't be surprised if it wasn't these delays that cause issues with the game itself.

Attached File(s)

•  eduke32.log (9.1K)
  Number of downloads: 588
•  eduke32_or_mapster32.log (178bytes)
  Number of downloads: 583

[Hendricks266]

That may actually be it... This is the most unhelpful crash log I've ever seen. (Not anything you did wrong, Paul, just the laptop.)

Failed to init bfd from (C:\Windows\system32\ig4icd32.dll): 1 1 0
0x10593299 : C:\Windows\system32\ig4icd32.dll : DrvPresentBuffers 
Sun Mar 31 11:15:26 2013
---------------

Though, ig4icd32.dll is not normally part of EDuke32's execution so I would venture it is some problem outside our scope.

[Helixhorned]

Hendricks266, on 31 March 2013 - 11:11 AM, said:

Though, ig4icd32.dll is not normally part of EDuke32's execution so I would venture it is some problem outside our scope.

Intel Graphics drivers :-/.

Forge, why did you remove the attachements? I still couldn't reproduce your bugs, but IIRC the crash log looked similar to Paul's -- one line somewhere not in our code, or no useful line at all. Were you running the debug version?

[Forge]

you couldn't reproduce it on your end so i figured i'd just do what i can to avoid it on my end.

0x613bab : C:\Duke3D\eduke32.debug.exe : pseudo-reloc-list.c
Fri Mar 29 16:27:50 2013
---------------

[Paul B]

Thanks for looking into this for me, definitely reassuring knowing its outside of Eduke. What's interesting about the display driver is I noticed its detecting an Intel driver to run Eduke and the Device Manager indicates two types of video drivers are installed. Intel HD Graphics and Ati Radeon HD 6470. But I have a feeling that the intel driver is not the correct driver as the laptop has a Radeon Graphics sticker on it. For some reason this laptop thinks its an intel graphics when really its an ATI processor. Very Strange.

[Plagman]

Most notebooks these days are hybrid, they have an integrated Intel chip and an ATI/NVIDIA chip that's used only for specific apps. You might need to launch EDuke32 with a special context menu option or add it to a list of OK apps in the ATI control panel.

[LeoD]

Mapster64 crash (self compiled, crashes every time).
Btw., I think ebacktrace1*.dll needs to be rebuilt in order to get the new crash.log extension.
Building ebacktrace1-64.dll fails for me, there seems to be something wrong with my include path.

Attached File(s)

•  eduke32_or_mapster32.crash.log (1.93K)
  Number of downloads: 642
•  mapster32.log (636bytes)
  Number of downloads: 629

[Hendricks266]

LeoD, on 02 April 2013 - 03:52 PM, said:

Btw., I think ebacktrace1*.dll needs to be rebuilt in order to get the new crash.log extension.
Building ebacktrace1-64.dll fails for me, there seems to be something wrong with my include path.

Yes.

Try the instructions just above "compiling".

[Hendricks266]

LeoD, on 02 April 2013 - 03:52 PM, said:

Mapster64 crash (self compiled, crashes every time).

This crash had less to do with 64-bit architecture and more to do with your extra-long custom version string.

Until I commit the fix, you can fix it in your local copy by deleting the line "char tempbuf[16];" from build.c.

[LeoD]

Hendricks266, on 05 April 2013 - 02:50 PM, said:

This crash had less to do with 64-bit architecture and more to do with your extra-long custom version string.

I have to apologize because I should have guessed that on my own. It once happened when I compiled for Raspbian as well, but I did not bother to go after it back then.

[LeoD]

Trying to switch from Polymer HRP to software renderer:

Attached File(s)

•  eduke32_or_mapster32.crash.log (3.18K)
  Number of downloads: 849
•  eduke32.log (73.47K)
  Number of downloads: 498

[Hendricks266]

Could you test if r3670 and r3671 work? r3670 should address this issue. If so, nice catch! That's exactly the kind of bug we need to weed out of winlayer before 64-bit builds can be considered stable.

[LeoD]

Hendricks266, on 12 April 2013 - 08:07 PM, said:

Could you test if r3670 and r3671 work? r3670 should address this issue.

I couldn't reproduce the crash with r3466. It hasn't occurred since using newer versions.

[LeoD]

Not easy to reproduce, but possible ... EDuke64 gives no crashlog, but only a SIGSEGV message in the normal log.

Attached File(s)

•  eduke32_or_mapster32.crash.log (2.27K)
  Number of downloads: 553
•  eduke32.log (2.22K)
  Number of downloads: 542
•  eduke64.log (3.38K)
  Number of downloads: 530

[Helixhorned]

LeoD, on 18 April 2013 - 02:35 PM, said:

Not easy to reproduce, but possible ... EDuke64 gives no crashlog, but only a SIGSEGV message in the normal log.

Looks like a light with an invalid sector number somehow gets processed by Polymer.

But... are you deliberately running the netcode to stress-test it? Does it also happen in an SP game? If not, I think these reports should be addressed to Jasper or TX. I don't know how they split the work, but the code badness I reported here is still not fixed.

[LeoD]

Helixhorned, on 19 April 2013 - 02:37 AM, said:

Looks like a light with an invalid sector number somehow gets processed by Polymer.

But... are you deliberately running the netcode to stress-test it? Does it also happen in an SP game? If not, I think these reports should be addressed to Jasper or TX. I don't know how they split the work, but the code badness I reported here is still not fixed.

I'm play-testing maphacks for multiplayer pickups and the likes. Hasn't happened it SP games yet. Net code messing up Polymer rendering gives me a pretty bad feeling...

[LeoD]

Hendricks266, on 12 April 2013 - 08:07 PM, said:

Could you test if r3670 and r3671 work? r3670 should address this issue. If so, nice catch! That's exactly the kind of bug we need to weed out of winlayer before 64-bit builds can be considered stable.

The crash has returned (mostly the same crashlog), I'll try to reproduce it tomorrow.

Attached File(s)

•  eduke32.log (9.15K)
  Number of downloads: 507
•  eduke32_or_mapster32.crash.log (3.18K)
  Number of downloads: 527

[LeoD]

LeoD, on 25 April 2013 - 02:18 PM, said:

The crash has returned (mostly the same crashlog), I'll try to reproduce it tomorrow.

Savegame for 65bit exe: load the game and use the menu to switch from polymer to software renderer.

Attached File(s)

•  Crash64-esv-log-r3723.zip (465.64K)
  Number of downloads: 466

[LeoD]

Down to 64bit again ... Mapster64 crash (reproducible)

Attached File(s)

•  eduke32_or_mapster32.crash.log (1.49K)
  Number of downloads: 503
•  mapster32.log (2.51K)
  Number of downloads: 478

[Helixhorned]

LeoD, on 04 May 2013 - 06:13 AM, said:

Down to 64bit again ... Mapster64 crash (reproducible)

You mean it just crashes with any map, or a particular one? No problems on a Linux/64 build here.

[LeoD]

Helixhorned, on 04 May 2013 - 06:40 AM, said:

You mean it just crashes with any map, or a particular one? No problems on a Linux/64 build here.

Win64, particular map/mod. Drop this ZIP into autoload, load 16384.map into mapster64.exe and zoom out. (Move away any other DEFs and CONs.)

[Helixhorned]

Sorry, I don't debug binaries built by others. The crash doesn't happen here, and classic rendering routines (where it is happening for you) aren't OS-specific, so it should behave the same with a recent EDuke32 build on Win64. I think Hendricks266 build that particular exe? What revision is it based on? What compilation flags were used?

Edit: moreover, I'm not getting any suspicious messages or aborts from a Clang build with address and undefined behavior sanitizers enabled (which do a fair amount of checking).

[LeoD]

Trying to load Earth Base ...

Attached File(s)

•  eduke32_or_mapster32.crash.log (2.56K)
  Number of downloads: 466
•  eduke32.log (1.78K)
  Number of downloads: 491

[LeoD]

Mapster32 zoom crash. Load BobSPX, go to sprite 965 (seenine) @ 29696/-99328 and zoom in to the Min...

Attached File(s)

•  eduke32_or_mapster32.crash.log (1.96K)
  Number of downloads: 464
•  mapster32.log (1.55K)
  Number of downloads: 480

[Helixhorned]

LeoD, on 09 May 2013 - 10:12 AM, said:

Trying to load Earth Base ...

This map is extremely corrupt: some wall's .nextsector/.nextwall members are out of the MAX* bounds, so it will report corruption level 5 with the new Mapster32 build. I've attached a map where the problematic walls have been made white. In the ZIP, there's also a list of their wall numbers. Use [']+[J] / [W] in Mapster32 to jump to a wall given by its index.

LeoD, on 09 May 2013 - 02:24 PM, said:

Mapster32 zoom crash. Load BobSPX, go to sprite 965 (seenine) @ 29696/-99328 and zoom in to the Min...

This is a duplicate of the crash you reported above.

Attached File(s)

•  EBase_fix.zip (125.08K)
  Number of downloads: 442

[LeoD]

Helixhorned, on 10 May 2013 - 04:54 AM, said:

This map is extremely corrupt: some wall's .nextsector/.nextwall members are out of the MAX* bounds, so it will report corruption level 5 with the new Mapster32 build. I've attached a map where the problematic walls have been made white. In the ZIP, there's also a list of their wall numbers. Use [']+[J] / [W] in Mapster32 to jump to a wall given by its index.

Thanks!

Helixhorned, on 10 May 2013 - 04:54 AM, said:

This is a duplicate of the crash you reported above.

Didn't r3188 fix that first zoom issue?