[americantabloid]

Hi!

When I go thru my media files, I sometimes experience this bug,

R6025 - pure virtual function call

Here is a screenshot of the errortext,

http://i475.photobucket.com/albums/rr117...nction.jpg

I was scrolling thru my media files, first my video directory, then my series directory, doing some fanart updates and thumbs updates on my network share when it happened.

System specs, XBMC Atlantis 4, Vista x64, Radeon x1950 pro, Catalyst 8.8.

I knew I read on the forum about a similar problem with the pure virtual function error, but I have searched on pure, pure virtual, r6025 and didn't find it.

Xbmc debug log
http://pastebin.com/f400e62fc

I also know that there where a similar ticket created on trac, but I can't seam to find that either...I know that you asked for more info in that ticket and thats why I created this thread.

Please let me know if you want any additional information or help with testing done.

Best regards
at

[jmarshall]

The key is to get it reproducible. It appears as though it's only in some circumstances, which is why I can't reproduce it. Once I can reproduce it (reliably) then I can investigate.

The ticket is here:

http://trac.xbmc.org/ticket/4520

Feel free to reopen it and add your experience.

Cheers,
Jonathan

[americantabloid]

Hi Jonathan!
Will try some more during the week and try to see if I can find any common causes for the bug to show up, so I can get it reproducible for you.

Thanks for the link.

Regards
at

[xeonicxpression]

The key is to get it reproducible. It appears as though it's only in some circumstances, which is why I can't reproduce it. Once I can reproduce it (reliably) then I can investigate.

The ticket is here:

http://trac.xbmc.org/ticket/4520

Feel free to reopen it and add your experience.

Cheers,
Jonathan

Sorry that I've kind of been MIA since creating that ticket. Hopefully I'll be able to tackle it again this week. My computer has been out of commission for quite awhile now. Life got in the way of getting it up and going again. I'll upgrade to the latest atlantis release and let you know if it still happens.

[elupus]

http://support.microsoft.com/kb/125749

might be of help.

[jmarshall]

elupus:

It appears to be due to two simultaneous accesses to the imagelib dll, but I'm not sure how it's occuring. The grabbing of the dll is critical sectioned off, but perhaps it's due to the ResolveExports() call not being protected, and the dll.Load() function being called almost simulataneously from different threads??

As can be seen from the Debug Log, the exception occurs when the dll is being unloaded.

Wish I could reproduce it so that I can get a callstack!

Cheers,
Jonathan

[americantabloid]

Hi Jonathan!

I have been doing some additional testing.

I have run XBMC on my laptop, connected to the same share as on the Vista64 machine that has resulted in the R6025 error.

Much to my surprise I can't get the error to reproduce on my laptop.

The spec of the laptop is HP nx7000 Pentium M 1.5GHz, Radeon 9200 graphics (Hp supplied driver) and Windows Xp Professional SP3.

What I did to try reproduce the bug is move around my series libary, where I have the series sorted in file structure like series name\season\episodes.

(I think the bug has something to do about going thru the libary and let XBMC generate screenshots from the episodes.)

Then I did the same with my Vista machine and after a short while I run into the R6025 error. To bad I didn't start XBMC with the "Run as an Administrator" option that I need to get an XBMC debug log written.

Anyway I started again and after awhile XBMC hang again. When I Ctrl+Alt+Del to get back to the desktop, it said that XBMC had stopped working. (I did not get the prompt about the R6025, but I took a quick look at the new debug log and seamed similar like the previous posted log, so I decided to post it anyway.)

http://pastebin.com/f3ca61368

Could the bug have something to do about running on dualcore vs single core machines?

Best regards
at

[jmarshall]

It could be affected by number of cores, and it is certainly the same problem.

[americantabloid]

Hi again!

Would it help if you could have Remote Desktop access to my machine and use it for debug purpose?

I don't know how to get you any more useful information.

Regards
at

[jmarshall]

Nope. The only thing that will make a difference is finding out how to reproduce it reliably (i.e. every time, or at least close to it).

[americantabloid]

Did a test to run XBMC on my computer that I use for fileshare duties. It has a Pentium E2160 processor (dual core), Windows Xp Pro SP2 and tried usual procedure, updating information of the media files, making XBMC generate thumbs when there is non to download for series. Just doing the same as on the Vista machine, same files since it is on the file sharing machine.

Can't get it to reproduce on this machine. Will do clean install on the Vista machine and see if it helps.

//at

[Freddo]

I'm having a problem with my new skin, I think this is related to the use or the largeimage control, and ImageLib.dll, basically any time an image is being loaded using largeimage theres a chance XBMC will crash, it seems to be quite random what'll set it off, sometimes it takes hours sometimes it happens right away, but the more largeimage loads the more likely it is to happen.

http://pastebin.com/f70acab6d

There's the debug log, you can see at the very end it freaks out due to msvcr71.dll.

Build being used is atlantis beta1, although I've also encountered the same problem on the latest SVN and various builds in between.

Platform is windows vista ultimate, specs are well above requirements, bar the intel gma945 graphics but to my knowledge that shouldn't affect gui usage (I'm just using this machine as a skinning tool basically).

So, can a Dev take a look at the log and tell me, is this something I'm doing wrong with my skin or is it an actual bug? if so I'll post a ticket for it.

[WiSo]

I don't think it's related to MSVCP71.dll as we don't need since we emulate it.
I assume it's something in imagelib maybe jmarshall can give a hint here?

[jmarshall]

There's a trac report already for this. We believe it's due to the imagelib being loaded/unloaded in a not-threadsafe manner (I suspect due to the resolving of imports/exports)

http://trac.xbmc.org/ticket/4520

Please add your findings to that. The key is to get it reproducible (easily) for a dev so we can hit the callstack to see what is possibly going wrong.

The log there is slightly different, but it gave the same error, so may well be another manifestation of the same problem.

Cheers,
Jonathan

[Freddo]

appended my findings but I don't have permissions to re-open the ticket. I believe this can be made reproducable by simply making extensive use of the largeimage control in a skin, since it seems to have started occurring as soon as I started using them.