Please provide the module that crashes OpenMPT.

it will be difficult.
it's a 300 mo module and i have a 63 kbs connection ....
(but bigger modules don't crash)

Seems like I won't be able to use a crash dump from that version as I don't have the debugging files from the buildbot builds, so please download the build from http://sagagames.de/stuff/mptrack.exe - reproduce the crash with that one and upload the generated memory dump.

Please add a comment the next time you add a crash dump, I didn't see the uploaded file, so once again I don't have the debug information for this test build anymore. A new build is already available under the same address, for which I have debug information.

oops sorry.
anyway it also crashes with xp.
I think there was a problem during my backup.
I remade another same module and I have no problem.
thank you jojo.

So the bug is gone or what?

only one module crashed in vista and xp, not the other ones.
so i think it was an exceptional backup problem only with this module.
I remade exactly the same module and i have no more problem with.

Well, of course it's possible that some bad disk sectors or whatever have caused the crash, but the disassembly from the crash dump doesn't look like that (though we had an issue like that before). So if you still have the file which caused the problem, please recreate the problem using the latest test version, just to go sure. We don't want OpenMPT to crash on valid or corrupt files.

so.
with the just now downloaded:
OpenMPT 1.22.07.22-r3726!+ TEST
Build Date: 2014-02-17 01:37:57
and exactly the same crashed backuped module.

2014-02-17 13.26.08.7z uploaded

Thanks, I see what's most likely going on there now; I'd guess that the file got truncated when saving / copying, so when loading some sample, it's reading past the file end. Should be able to fix that easily.

manx, this was introduced in your refactoring commit 2438, the security checks in CopyStereoSplitSample were removed. Previously, if there was not enough incoming data, the method read as much of the left and right channel as possible, but now it simply crashes if there's not enough incoming data for the right channel. Can you fix this?

This should be fixed in r3727 now, could you please check if http://manx.datengang.de/openmpt/temp/mptrack-3727.7z works without crashing now?
Keep in mind though that the module that crashed before is most likely corrupted somehow and has missing sample data for at least one of its samples.

fixed with r3727.
no problem to open the module.
one truncated sample.
some missing samples.
thanks manx.