CCSv5.3 crashes when lanching debug session if I use the -clean option

Hi Juan,

could you please enable logging and send us any error logs and dump files (.dmp) generated?  That would be most helpful digging into the issue.  THis wiki shows how.

http://processors.wiki.ti.com/index.php/Troubleshooting_CCS

Best Regards,
Lisa

Hi Lisa,

I attach the log file located in .metadata folder 4571.logfile.log

I couldn't find any .dmp file. I did a windows search for files with the *.dmp file extension at the root of the CCStudio install directory but no matches were found. I think  it makes sense because the "crash  window" that ususally appears when CSS is invoked after crashing didn't show up. I guess this behaviour is not usual.

Please, do not hesitate to contact me again if you need additional files or info.

Best,

Hi Juan,

that is unfortunately simply a workspace log.   As per the wiki, we need the hs_err log files or any .dmp files generated.  Would you be able to send those?

Best Regards,
Lisa

Hi Lisa,

Sorry for attaching the wrong .log file. I've tried to reproduce the problem and collect the diagnostic info as detailed at the troubleshooting CCSv5 guide.

- 1. I invoke CCSv5 with the '-clean' option.

- 2. I enable all the diagnostic log files as showed in the snapshot below.

3. I start a debug session. This time, the session runs well but when I stop it, then CCS hangs.

4. I kill the CSS process (Ctl+alt+del). 

5. I check the diagnostic files.

  - Debug Server Log file was created but I cannot atach it because I received a error message when I try to upload it. The reason would be the file size (~20 MB). Find the URL link attached here 

 - No Debug Server Dump file was created (folder is empty)

I afraid I don't know anything about the hs_err log files and there's no info about how to find them in the troubleshooting guide. I did a windows search for  the regular expression 'hs_err*' but no matches were found.

Hi Juan,

could you please do a couple things

- please send us the debug server log

- please double check for any hs_err files and send those

These will help start to give us an idea of where the crash might be occuring.

Best Regards,

Lisa

Hi Lisa,

I updated my previous post before reading your new post. Please, find attached there the Debug Server Log file. Regarding the hs_err files, I did a window search for the regular expression '*hs_err*' but no matches were found. I have no idea about such files so could you tell me how to generate and find them?

Regards,

Hi Juan,

the hs_err log files should be in the .TI folder ... I would double check you are also showing hidden files.   A simple way as well is to collect them from the help->ccs support menu (there is an archive logs button to help collect everything)

If you still have problems let us know but hopefully between the wiki and this we should get there.  you do need to capture a crash.  It does seem you are on the right track though wtih the screenshot you sent.

Best Regards,
Lisa

Hi again Lisa,

I double checked the content of .dmp folder (including the hidden files) but it is still empty. I also tried the option help->css support menu -> archive logs (after enabling the Debug Server log) but the .zip file didn't contain any .dump or hs_err file. As I told you in my previous posts, CCS didn't show up the error reporting tool window after crashing. I guess it is due to kill the ccs process abruptly. I don't know if it could be the reason because those log files are missing. Anyway, I will be alert for the next time the error window appear and I will check the log files then.

Best Regards,

Juan

Hi Juan,

yes, please send us logs as soon as you are able to capture a crash.

Best Regards,
Lisa

Hi Lisa,

Finally, I was able to capture the crash. Find attached the .dmp files 5722.log_files.zip

Sincerely, Juan.

Juan,

Your initial post indicated that CCS would crash every time if you used the -clean option.  However, later on in this thread I get the impression that you were unable to get it to crash and subsequently find a .dmp file.  So where are these attached .dmp files coming from?  Are you no longer running with the -clean option and the attached .dmp files are from some other crash, or did does it not actually crash very often when using -clean and these are the .dmp files associated with that crash?  Not every crash will generate a .dmp file, so it's possible that the crash associated with -clean simply doesn't generate one.

I ask because the .dmp files point to a known crash in 5.3 (that's fixed in 5.4) which occurs when the java garbage collector runs at just the right time when unwinding the callstack.  Since that crash is a race condition, I would expect you to see it only infrequently, and I would not expect it to be associated with using the -clean option.

Darian

Hi Darian,

I will try to be as clear as possible. I will detail the sequence of actions that I carried out 

1. I launch CSS with -clean option

2. I start a debug session and CCS hangs. I kill the CCS process.

3. I relaunch CCS with -clean option and a diagnostic error window pop up (it usually doesn't  happen after CCS crashes as indicated in action 2)

4. I check the .dmp  folder and I find the files that I attached in my previous post. 

As I wrote in the previous posts, when the diagnostic error window doesn't appear (assuming that CCS process was killed earlier), then the .dmp folder is empty.

Juan,

I believe I understand what happened when the log file was collected.  My confusion is that you originally stated that CCS crashed, but when you collected the log file it didn't crash - it hung and you had to kill it. 

So is it no longer crashing and is now hanging, or does enabling logging seem to change the crash into a hang?

Darian

Sorry Darian, it is my fault.

From the beginning, the problem was that CSS hung. I didn't think that the difference was important for you. It was an abuse of language.

Regards, Juan.

Juan,

No worries, but that explains the dump files.  The ones you attached were from the known occasional crash I mentioned before and have nothing to do with this hang.

Now, the logs you submitted show that you were able to launch and terminate.  Does it normally hang after terminating, or is that just what happened when logging was enabled?  If it normally hangs earlier, then it might not hang when logging is on because the extra time spent writing the file was enough avoid the race condition. 

Anyway, the next time it hangs, open up the task manager in windows, find the ccstudio.exe process, right click on it and select "Create dump file".  I believe the file will be rather large, so if it doesn't zip up small enough to attach to this thread, please let me know and I'll setup a temporary ftp site that you can send it to.

Regards,

Darian

Hi Darian,

It hung regardless of the logging state. I will follow your piece of advice and collect the dump file the next time CCS hangs.

Sincerely, Juan.

Juan,

One more thing: It would be useful to know the jvm stacks too (depends if the hang is occurring in eclipse itself or if it's occurring in our debugger).  So as well as collecting the task manager dump file, start running CCS from the command prompt using “eclipsec --launcher.ini ccstudio.ini --clean > threadDump.log”. After the deadlock occurs, activate the cmd prompt window where eclipsec was started from and press CTRL+Break. This should dump the thread stack to threadDump.log file.  Please attach that file as well.

Thanks,

Darian