32-bit version crashes due to alignment issues

[Richard]

I get the message
dc++ has stopped working

As soon as I start DC++

If I get it working for a few minutes it closes itself anyway.

Running the latest downloadable DCPlusPlus-0.862
OS is Windows 10.

What am I doing wrong?

Richard

[eMTee (realprogger)]

Thank you for the report.
Probably you do nothing wrong just encountered a fatal crash bug. You might try https://answers.launchpad.net/dcplusplus/+faq/337 to provide us more information about this crash in order to get a solution.

[Richard (ricleh)]

Ok

Is this what you wanted?

DC++ has crashed on 2016-10-31 at 17:05:01.
Please report this data to the DC++ team for further investigation.

DC++ version: DC++ v0.862 ("ae26d9b8b3b4 - 2016-10-02 18:40 +0200")
TTH: OP74BZ43HI6QS36W2SAJ7CXZ4G6VUUXGWFRO6JI
Compiled with MinGW-w64's GCC 6.2.0
Exception code: c0000005
Windows version: major = 6, minor = 2, build = 9200, SP = 0, type = 1
Processors: 2 * x86
System memory installed: 2,93 GiB
Writing the stack trace...

DCPlusPlus: H:\Dev\DC++\repo/dcpp/FinishedManager.cpp (169), function:
void/unknown onComplete(FinishedManager* const this, Transfer* t, bool
upload, bool crc32Checked)
DCPlusPlus: H:\Dev\DC++\repo/dcpp/FinishedManager.cpp (265), function:
void/unknown on(FinishedManager* const this, Upload* u)
DCPlusPlus: H:\Dev\DC++\repo/dcpp/Speaker.h (47), function: void/unknown
on(UploadManager* const this, UserConnection* aSource, Upload* u)
DCPlusPlus: H:\Dev\DC++\repo/dcpp/Speaker.h (47), function: void/unknown
on(UserConnection* const this)
DCPlusPlus: H:\Dev\DC++\repo/dcpp/FilteredFile.h (192), function: size_t
read(FilteredInputStream<dcpp::ZFilter, true>* const this, void/unknown
rbuf, size_t& const len, uint8_t* rb, size_t totalRead, size_t
totalProduced)
DCPlusPlus:
H:/MinGW/lib/gcc/i686-w64-mingw32/6.2.0/include/c++/bits/stl_algobase.h
(368), function: vector<dcpp::BufferedSocketListener*,
std::allocator<dcpp::BufferedSocketListener*> >&
operator=(vector<dcpp::BufferedSocketListener*,
std::allocator<dcpp::BufferedSocketListener*> >* const this,
vector<dcpp::BufferedSocketListener*,
std::allocator<dcpp::BufferedSocketListener*> > const& const __x)
DCPlusPlus: H:\Dev\DC++\repo/dcpp/Speaker.h (47), function: void/unknown
threadSendFile(BufferedSocket* const this, InputStream* file, size_t
sockSize, size_t bufSize, ByteVector readBuf, ByteVector writeBuf,
size_t readPos, bool readDone)
DCPlusPlus:
H:/MinGW/lib/gcc/i686-w64-mingw32/6.2.0/include/c++/bits/unique_ptr.h
(76), function: bool checkEvents(BufferedSocket* const this)
DCPlusPlus: H:\Dev\DC++\repo/dcpp/BufferedSocket.cpp (499), function:
int run(BufferedSocket* const this)
DCPlusPlus: H:\Dev\DC++\repo/dcpp/Thread.h (110), function: DWORD
starter(void/unknown p, Thread* t)
ntdll: [Failed to load the debugging data into memory (error: 2)]
RtlClearAllBits
ntdll: [Failed to load the debugging data into memory (error: 2)]
RtlClearAllBits

Information about the crash has been written.

Richard

Den 2016-10-31 kl. 16:37, skrev eMTee:
> Thank you for the report.
> Probably you do nothing wrong just encountered a fatal crash bug. You might try https://answers.launchpad.net/dcplusplus/+faq/337 to provide us more information about this crash in order to get a solution.
>

[eMTee (realprogger)]

Yes this is the one we wanted.
If you experience more crashes, could you please check time to time if the content of the saved crashlog is changing or stays the same?
It looks to be a crash after an upload of a (chunk of a) file anyway.

[Richard (ricleh)]

Ok

Sure I can do that. Got this one from the last crash and it´s not identical.

DC++ has crashed on 2016-11-01 at 09:59:01.
Please report this data to the DC++ team for further investigation.

DC++ version: DC++ v0.862 ("ae26d9b8b3b4 - 2016-10-02 18:40 +0200")
TTH: OP74BZ43HI6QS36W2SAJ7CXZ4G6VUUXGWFRO6JI
Compiled with MinGW-w64's GCC 6.2.0
Exception code: c0000005
Windows version: major = 6, minor = 2, build = 9200, SP = 0, type = 1
Processors: 2 * x86
System memory installed: 2,93 GiB
Writing the stack trace...

DCPlusPlus: H:\Dev\DC++\repo/dcpp/QueueItem.cpp (105), function: Segment
getNextSegment(QueueItem const* const this, int64_t blockSize, int64_t
wantedSize, double donePart, int64_t targetSize, int64_t start, int64_t
curSize)

Information about the crash has been written.

Richard

Den 2016-11-01 kl. 15:08, skrev eMTee:
> Yes this is the one we wanted.
> If you experience more crashes, could you please check time to time if the content of the saved crashlog is changing or stays the same?
> It looks to be a crash after an upload of a (chunk of a) file anyway.
>
> ** Changed in: dcplusplus
> Status: Incomplete => Confirmed
>
> ** Summary changed:
>
> - dc++ has stopped working message at start
> + Crash after an upload transfer has finished (DC++ 0.862)
>
> ** Changed in: dcplusplus
> Importance: Undecided => High
>

[eMTee (realprogger)]

Were there any changes in your setup (upgrades etc...) after these massive amount of crashes started to appear?

[Richard (ricleh)]

They appeared after I updated DC++

Nothing else has changed

Richard

Den 2016-11-01 kl. 16:28, skrev eMTee:
> Were there any changes in your setup (upgrades etc...) after these
> massive amount of crashes started to appear?
>

[eMTee (realprogger)]

Any reason why do you use the 32-bit version of DC++? What CPU does your computer have?
Note that from version 0.860, DC++ requires more modern (about or less than 10 year old) CPUs to run.

[Richard (ricleh)]

Hello

The computer is

Am I using the wron version?

Richard

Den 2016-11-01 kl. 20:32, skrev eMTee:
> Any reason why do you use the 32-bit version of DC++? What CPU does your computer have?
> Note that from version 0.860, DC++ requires more modern (about or less than 10 year old) CPUs to run.
>

[eMTee (realprogger)]

No you're using the right one since from your screenshot it's clear that you have a 32-bit operating system, therefore you can use the 32-bit version of DC++ only.
Your CPU, while not too new, still must be eligible so that's not the issue here.

Until other devs come up with more ideas I guess to be able to actually use the program, it's best for you to downgrade to the DC++ version that was stable (presumably it was 0.851, available here:
 https://sourceforge.net/projects/dcplusplus/files/DC%2B%2B%200.851/DCPlusPlus-0.851.exe/download )

If you decide to do so please confirm that the crashes stopped with the older version.

[Richard (ricleh)]

Ok

Thankyou for your patience.

Downloaded and installed now.

Let you know how it works.

Richard

Den 2016-11-01 kl. 21:50, skrev eMTee:
> No you're using the right one since from your screenshot it's clear that you have a 32-bit operating system, therefore you can use the 32-bit version of DC++ only.
> Your CPU, while not too new, still must be eligible so that's not the issue here.
>
> Until other devs come up with more ideas I guess to be able to actually use the program, it's best for you to downgrade to the DC++ version that was stable (presumably it was 0.851, available here:
> https://sourceforge.net/projects/dcplusplus/files/DC%2B%2B%200.851/DCPlusPlus-0.851.exe/download )
>
> If you decide to do so please confirm that the crashes stopped with the
> older version.
>

[Richard (ricleh)]

Hi

Working so far with the downgrade!

Thanks

Den 2016-11-01 kl. 21:50, skrev eMTee:
> No you're using the right one since from your screenshot it's clear that you have a 32-bit operating system, therefore you can use the 32-bit version of DC++ only.
> Your CPU, while not too new, still must be eligible so that's not the issue here.
>
> Until other devs come up with more ideas I guess to be able to actually use the program, it's best for you to downgrade to the DC++ version that was stable (presumably it was 0.851, available here:
> https://sourceforge.net/projects/dcplusplus/files/DC%2B%2B%200.851/DCPlusPlus-0.851.exe/download )
>
> If you decide to do so please confirm that the crashes stopped with the
> older version.
>

[eMTee (realprogger)]

Thank you for the info!

Let us know if you're up for testing a special build of DC++ that might help us to pinpoint the issue.

[poy (poy)]

Hi,

I know this is asking for a lot on your side, but since the crash log doesn't tell us much, could you try the 2 versions in-between 0.851 & 0.862? You can see them on <https://sourceforge.net/projects/dcplusplus/files/>; take the .exe file to get the automated installer.

If you get crash reports on these versions, it would be useful to get them too.
If you don't, that would be good to know as well, so we can reduce the amount of potential changes that may have caused this.

Thanks

[eMTee (realprogger)]

@poy, cologic suggests https://gcc.gnu.org/bugzilla/show_bug.cgi?id=40838#c63 might be the issue here so I've done a build of the 0.862 release rev. with -mstackrealign for a possible further test.
However, while none of us tried a 32-bit build of 0.86x on a 32-bit OS before the release, I have run an SSE2 optimized 32-bit build on my Win7 x64 for a prolonged time before the release of 0.861, just to see if there are any issues.
Also, I have just tried 0.862 on a 32-bit Win7 in a VM and there were no crashes with multiple threads running.
Now the only difference between my test scenarios and the reporter's setup is the OS. Unfortunaltely MS has only 64-bit Win10 test VMs available so I couldn't have tried it on Win10 32-bit yet.

[Richard (ricleh)]

Anything I should do to help?

Richard

Den 2016-11-05 kl. 11:33, skrev eMTee:
> @poy, cologic suggests https://gcc.gnu.org/bugzilla/show_bug.cgi?id=40838#c63 might be the issue here so I've done a build of the 0.862 release rev. with -mstackrealign for a possible further test.
> However, while none of us tried a 32-bit build of 0.86x on a 32-bit OS before the release, I have run an SSE2 optimized 32-bit build on my Win7 x64 for a prolonged time before the release of 0.861, just to see if there are any issues.
> Also, I have just tried 0.862 on a 32-bit Win7 in a VM and there were no crashes with multiple threads running.
> Now the only difference between my test scenarios and the reporter's setup is the OS. Unfortunaltely MS has only 64-bit Win10 test VMs available so I couldn't have tried it on Win10 32-bit yet.
>
> ** Bug watch added: GCC Bugzilla #40838
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=40838
>

[Richard (ricleh)]

Hi

I´l give a try.

Richard

Den 2016-11-05 kl. 10:03, skrev poy:
> Hi,
>
> I know this is asking for a lot on your side, but since the crash log
> doesn't tell us much, could you try the 2 versions in-between 0.851 &
> 0.862? You can see them on
> <https://sourceforge.net/projects/dcplusplus/files/>; take the .exe file
> to get the automated installer.
>
> If you get crash reports on these versions, it would be useful to get them too.
> If you don't, that would be good to know as well, so we can reduce the amount of potential changes that may have caused this.
>
> Thanks
>

[eMTee (realprogger)]

I can confirm this issue with 0.862 and also with the head rev. on Win10 x86 running on a VM (but not on Win7 x86).

[Richard (ricleh)]

Hi again

Tried the first one a couple of days and it´s not crashing.

Upgrading later today to the next one you wanted me to try.

I´ll get back on that after a couple of days

Richard

Den 2016-11-05 kl. 10:03, skrev poy:
> Hi,
>
> I know this is asking for a lot on your side, but since the crash log
> doesn't tell us much, could you try the 2 versions in-between 0.851 &
> 0.862? You can see them on
> <https://sourceforge.net/projects/dcplusplus/files/>; take the .exe file
> to get the automated installer.
>
> If you get crash reports on these versions, it would be useful to get them too.
> If you don't, that would be good to know as well, so we can reduce the amount of potential changes that may have caused this.
>
> Thanks
>

[poy (poy)]

Hey,
Don't bother with these tests - eMTee has been able to reproduce the issue and we are closing in on a fix.

[eMTee (realprogger)]

Yeah and I can also confirm this issue on Win 8.1 x86. It crashes with the log in comment #2 at every closal of the program if transfers are running.

[Richard (ricleh)]

Hi

Ok

Doing fine so far

Thanx

Den 2016-11-10 kl. 16:14, skrev poy:
> Hey,
> Don't bother with these tests - eMTee has been able to reproduce the issue and we are closing in on a fix.
>

[poy (poy)]

Fixed in DC++ 0.863.

[Richard (ricleh)]

Great work

I´ll try it out as soon as I can

Den 2016-11-12 kl. 01:33, skrev poy:
> Fixed in DC++ 0.863.
>
> ** Changed in: dcplusplus
> Status: Fix Committed => Fix Released
>