XDS510USB Plus DWARF information is corrupted

Hi Nelson - I've not heard of anything like this before.  I am going to move this to the Code Composer forum as they deal a lot more with the emulators and CCS.

-Lori

I've never seen this issue before but I did find another forum thread with the same error. Could you take a look at that one and see if it helps?

Other things you can try to rule out problems that may be coming purely from a CCS perspective would be the tips in General IDE and Debugger sections of the CCSv5 troubleshooting page.

Are you working with your own custom board? If so, it might also be worth a shot to contact Spectrum Digital about this since you have found that XDS510USB (non-Plus) emulators work fine with your boards but XDS510USB Plus exhibit this issue. They might be able to give some clues on key differences between the emulators that might cause issues during programming.

Here is the response I got back from Spectrum Digital:

The "DWARF information is corrupt" message is not related to the emulator or drivers. It is something related with the CCS compiler or loader. Take a look at the TI E2E forum for topics regarding the DWARF or post a message there and the TI engineers can help you further. If you want to try an emulator driver update, you can used the CCS updater to search for updated emulation driver.

Nelson,

A quick update on this.  We have been trying to reproduce this internally but have not been able to. 

If you are willing to send the associated .out file the next time this occurs it may help the team debug. 

Thank you

Lori

Hi Nelson,

I'm discussing this with the CCS debugger engineers and it seems that message is a bit of a catch-all for problems that occur when loading a .out file. To figure out what is going wrong, we'll need some more data. Would you please be able to do the following to capture some logs? 

The next time you encounter this error, go to the Help->CCS Support menu and turn on Debug server logging. Then repeat trying to load the same .out file without rebuilding (or hitting the green bug button) so that you get the same error.  You can do this by simply going to menu Run->Load->Load Program.

Once you have reproduced the problem with logging enabled, exit and restart CCS and try to load the same out file again without rebuilding.  You can do this by right-clicking on your target configuration file and launching selected configuration, then in the Debug perspective go to menu Run->Connect Target, and then Run->Load->Load Program. We want to know if just restarting CCS causes the issue to go away.

Please attach the generated debug server log file to this thread. Also, if the problem is consistently reproducible with the same outfile, could you send us the .out file as well?

Hi everyone

I expirienced the same issue today.
Specifying the DWARF version (select box in the Debug Options in the Compiler register) to "2" solved the problem for me. Maybe this is good for you too.

Regards
Joachim

I was finally able to get back to this issue.  Once I get this error message, we can usually get the next load to work (without rebuilding or closing the debugger).  Sometimes it takes 2-3 more attempts, but it will load correctly without rebuilding.  I am currently using the M36 dual-core blinky example, set the Flash configuration.  It is a lot harder to get the error message with this project, but it does happen occasionally.

Attached is the debug server log file.  The last information logged was the failure.

FYI: I did try Joachim's suggestion, and it did seem to help a little bit.  I didn't try setting the DWARF information to 2 in the data caputured above.

Joachim,

I tried this.  It seems to help some but I still get the error message.  It seems to be worse when the .out file is large as opposed to programming an example project .out file.

Nelson Norris said:

Attached is the debug server log file.

I do not see an attachment. Could you please reattach? Thanks.

2311.DebugServer2.zip

Thank you for the log file. Could you also please attach the .out file that generates this error for our development team to analyze?

Here is the .out file for the M3 core for the M36 Concerto.

1172.blinky_dc_m3.zip

Any updates?  I have also tried slowing the TCk rate down on the XDS510USB Plus to 10 and 12 MHz, but it still fails.

We have done some testing internally and believe this issue is now fixed in the Object File Server (OFS) dlls within CCS.

Would you mind installing CCS 5.4 beta and adding in some updated dlls and trying your program downloads to confirm if the issue is fixed?

To install the updates:
- download and install CCS 5.4 beta (5.4.0.00048) from http://processors.wiki.ti.com/index.php/Code_Composer_Studio_Beta_Downloads
- unzip the attached file ofs.zip
- replace the OFS dlls in the CCS install sub-directory \ccs_base\DebugServer\bin with these new dlls 

Try downloading/programming your executables with this new version and let us know if it still fails. 

These updates will be in the final 5.4 release.

7752.ofs.zip

Aarti,

Thanks for the update.  I will install the 5.4 beta and add the OFS dll files to see if that fixes our problem.  I will let you know of the results.

Nelson

Hi AartiG,

I also encounter the same issue, I installed the CCS5.4 and update the ofs dlls, but the load still fails. I attached my project, it's a hello world project running on TCI6638 cortex A15 cores. This project use the arm toolchain 5.0.3, and there is a boot.asm file in this project, if I exclude this file, the code can be loaded successfully, as long as I build the project with the boot.asm file, the load fails. The boot.asm file is from the arm toolchain 5.0.3 source code.6574.HelloWorld.zip.

Could you give some comments? Thank you very much.

Zhan.

Hi Zhan,

I have investigated your issue.

I find that in some case, CCS does not process the ARM BE debug info correctly.

I have fixed the bug and the fix will be included in CCS 5.4 official release.

Regards,

Raymond

Hi Raymond,

Could you let me know the official release plan? My customer system debug has been blocked by this issue, could you provide a patch for us first? Thank you very much.

Zhan.

Hi Zhan,

CCSv5.4.0 will be released on Apr 29 (next Monday).

If your customer needs a fix before that, please send me a msg.

Regards,

Raymond

Hi Nelson,

Just wondering if you had a chance to try out the patch update and if that helps resolve the error.

The problem reported by the other user Z.X was actually a different issue which we have now fixed also, but we would love to have confirmation from you whether you find the issue to be resolved in your environment or not.

I have installed the V5.4 beta with the patch and it appears to have corrected the issue.  I will verify once the official V5.4 release is done next week.  Thanks!

Nelson Norris said:

I have installed the V5.4 beta with the patch and it appears to have corrected the issue.  I will verify once the official V5.4 release is done next week.  Thanks!

Thank you very much for the feedback!

I have also tried the V5.4 beta and the patch.  It does not work for me.  About 2 out of 3 times, I get this error.  I have a large program that takes a while to program the flash (262K).  I looked for the official release, and it isn't there.  Any new ETA on a fix?  This is a big time-waster.

James lOCKWOOD said:

I have also tried the V5.4 beta and the patch.  It does not work for me.

The CCS 5.4 release is now expected end of next week.

What target are you working with? Can you attach your .out file here so we can check internally if the fix works with your .out file or not?

Will there be a difference between the RC on the beta download site and the official release of CCS 5.4?  If so, will a small download be able to update me when the release is official?

I am having the "DWARF" problem quite a bit on 5.3 right now and it is hurting my productivity.  

Cheers.

William,

Have you confirmed with 5.4 RC that the Dwarf corrupted error no longer occurs? Unless any show stopper bugs are identified in the RC, the final release should be the same as RC.

I have not tested 5.4 Beta, no.  I took a break from developing and moved to other tasks until the product is released.

I have installed (based upon my CCS "About" dialog screen) 5.4.0.00048, and installed the .dlls mentioned previously in this thread.  Is this the 5.4 RC, or is there another recent update I should try?

I do not have a makeable .out file at this time for the large project, and will be travelling all day tomorrow.  I can probably get it back to compiling on Thursday for you.

I have been working on a small 14K-ish sized test program for some RTSC modules I am writing, and have seen it give me the corrupt dwarf problem once with approximately 75 downloads to flash.  It seems to be size related.  The larger the file, the more likely you encounter the problem.

BTW - These two projects share just about nothing in common.  One is CppUTest, with the basic internal test programs they provide and a (CppUTest) platform configuration library I made for concerto.  The other tests a modified TimeStamp module, SystemGate, for xdc.runtime.gate which is decoupled from SYSBios so it doesn't bring it along the baggage.

Is the download link down?

This is my "out" file.  It was generated through XDCTools on the command line with my own build script, rather than from CCS, so it has the file extension of .xem3.

Good luck.

James

Update.  I Installed version 5.4 on another computer, and found same problems.  Realized I forgot to update the drivers.  After adding them, the problem seems to have gone away.  I am 99.9% sure I updated the drivers on the original computer I was having problems with.  Unfortunately, the original computer is on the other side of the country, and I won't be back for 2 weeks.

If you create a debug configuration and launch from there, on the target tab->Auto Run and Launch Options, check Continue debug launch if target connection or program load fails.  You can recover quickly without starting the lengthy download process all over again.  In the debug perspective, you can simply Run->Load->Load Symbols.  A good work-around that isn't nearly so painful.

James lOCKWOOD said:

Update.  I Installed version 5.4 on another computer, and found same problems.  Realized I forgot to update the drivers.  After adding them, the problem seems to have gone away

Are you referring to the final release of CCS 5.4(5.4.0.00091) that was posted today? I take it that the new version fixes the problem you have been using.

For the others that have posted on this thread with similar problems, the final CCS 5.4 release is now ready for download. Please download and install this version and let us know if you continue to see "corrupt dwarf" errors with this version.

No. CCS 5.4.00058.  I will download the released version and check it out.  I won't know for sure until I get back on my other computer in 2 weeks if the new release solves my issues I had back there.