Breakpoints unpredictable in R2011a

Since upgrading to R2011a, breakpoints in my code are unpredictable. Sometimes they work, and sometimes they don't. My code worked fine with breakpoints in R2010b. I don't have any "clear all" statements, and the breakpoints are still visible as red dots in my code. The code I'm using is on the MATLAB path, and there is only one copy.

Anyone else having difficulty with breakpoints in this latest release? Even better, anyone know how to fix it?

[Manicka]

I'm also having the same problem. Breakpoints worked fine with 2010b, but in 2011a Matlab skips them about half the time and then I need to set breakpoints at other lines as well to get the original breakpoints to work.

Have you tried reinstalling?

[TB]

I also have this problem. I am on Windows 7 x64.

[Vladimir]

I am experiencing the same problem on Windows 7 x64. Same code, same workflow, but issue appears on 2011a and not 2010b.

[SK]

Same problem here, also on Windows 7 x64. This problem did not appear before MATLAB R2011a.

[Matan Hofree]

Same on Linux x64 (CentOS release 5.6 (Final))

[Jan]

Did you ask the technical support already? What is the reply?

[Daniel Shub]

Can you post some code and instructions on which line to place the break point so we can try and replicate your problem? I am running r2011a and do think I have this problem.

[Friedrich]

This is a known issue and unfortunately not reproducible. MathWorks is currently investigating this further.

[Daniel Shub]

Is there a bug report on this? I love the known bugs that are being investigate that are not listed as bugs.

[Oleg Komarov]

This thing is starting to annoy me as well

What I do is:

click to clear > click to place the breakpoint back > save .m file

[Richard Crozier]

This is frankly, outrageous, yet true!!

Answer 1

[TB]

The official word I got from The Mathworks is that it is fixed in R2011b, but no workaround for R2011a.

[Jan]

Well, I do not like the method to fix bugs with new releases, because a new release will have new bugs and in consequence this strategy will not increase the total stability of Matlab. If I have to decide to buy a new Matlab version, I will not choose 2011a, because unpredictable breakpoints are very annoying. And I will not choose 2011b, because the currently unkown bugs might be even more annoying. And if the bugs are known, they will be fixed by choosing 2012a...

2010b, 2010a and 2009b had also serious bugs. I should buy an *old* Matlab version and choose 2009a. Not satisfying.

[Daniel Shub]

@Jan, I totally agree. I wish MATLAB would do something like a LTS. Something like r2009a would have service packs r2009a_sp_r2009b, r2009a_sp_r2010a, ... where the service packs would only fix bugs and come with some really good documentation about what changed.

[Friedrich]

EBR was created:

http://www.mathworks.com/support/bugreports/738717

[Jan]

Fine, now it is a documented bug. Nevertheless it remains a bug and will not be fixed although it concerns the elementary functionality of MATLAB 2011a. This reduces the value of this release dramatically.

[Robert]

Agreed. Suppose I only have 2011a, without money to buy a new release. Then I'm stuck with a buggy, useless program. If Mathworks sells me software with a bug, they should not then tell me, "Oh, by the way, if you want this to work correctly you have to pay even more."

[Daniel Shub]

While I am all for venting and in my opinion it is fine to vent here, nobody who can do anything is listening (not quite, but you get the idea). Go through the official channel of customer support to let them know how you feel.

Answer 2

[Alexander Kosenkov]

Extremely annoying bug, since R2011a. Breakpoints have their own lives, regardless of user actions. Just doesn't work.

As workaround, sometimes, it helps to modify file (add and remove space) and then saving it again. Otherwise I have to restart Matlab (obviously loosing all breakpoints). Clicking 'Clear all breakpoints in all files' and then setting breakpoints again - does not help for me.

Anyone here from Mathworks to comment?

[Robert]

This bug makes the R2011 versions useless to me. Maybe it's my own fault for writing software that needs debugging in the first place, but until the debugging feature works properly I'm sticking with R2010b.

Would love to know when the issue is fixed, if indeed it is a known issue as Friedrich claims above.

Answer 3

[Andreas Goser]

One user pointed me to this thread, as we were discussing aspects of the development timelines and fixing bugs. Indeed, MathWorks fixes bugs in the release under development and not by default for all affected older releases. But certainly, it is possible. And it happens. Please see the recent Service Pack releases. Those are responses on a number of (typically significant) bugs. In short, this is dependent on impact, frequeny and whether accpetable workarounds are available or not. I can just encourage you to describe the impact to Technical Support, so that they can provide this information to Development. You may think the impact is obvious? It is not.

[Jan]

Thanks for this comment.

I see the problem of the efficient communication of bugs. It is a big challenge for the technical support to filter the high impact problems from the noise of "EVAL does not work as expected" messages. Sometimes the support team is a little bit stubborn (if this term has a negative connotation: sorry, meant more neutrally - I'm not a native speaker), but they *never* left me alone with a problem until I confirmed that it is solved.

I think, this forum is helpful to improve the communication with the technical support: a discussion with other users helps to reduce the noise. If the bug-report contains a rough overview about the number of affected users (e.g. "<10" "<100" etc) and a tiny description of the underlying cause of the bug, it would be easier to estimate, if this is a strange exception or a general lack or quality. "This bug was fixed as of R2011b(7.13)" is not a very encouraging feedback of the feedback.

[Walter Roberson]

This is the first I have heard of the service pack releases ??

[Jan]

@Walter: Go to http://en.wikipedia.org/wiki/MATLAB and look for the "Release history". There are 2 SPs sincs 2006.

Answer 4

[Lucas García]

Hi Robert,

I haven't been able to reproduce this issue, but I have heard it from other users, so that makes it empirically a possible bug in R2011a. In any case, I will recommend that you report it to the Technical Support team mentioning your problem with as much detail as you can.

When you are debugging and type dbstatus, are the true breakpoint lines listed?

As a possible workaround, I will try the following:

>> feature accel off
>> feature jit off

[Jan]

It cannot be called "workaround" to disable the JIT acceleration. Some of the loops will need 100 times more processing time, such that a program can become completely useless even if it only concerns the debugging.

Answer 5

[Sean de Wolski]

Are you calling it in cell mode by clicking "Evaluate Cell"?

[Robert]

no

[Sean de Wolski]

Okay. Do you have multiple files of the same name and it's not running the one you have. E.g. same named files in different folder in which case the break point containing one might not be running?

which your_mfile_name

[Walter Roberson]

Also, is it possible that the file or directory is not writable? Breakpoints only work if the file is writable.

[Robert]

As I said, there is only one copy of this file.

And yes, the file is writable. If it was unwritable and that was causing problems, the breakpoints would never work. As I said, I'll hit run and they don't work, then I hit run again and they do work.

[Jan]

Any CLEAR ALL hidden in a subfunction?!

[SK]

Definitely no CLEAR ALL in my code. And the file is writable, too.

Answer 6

[Adam]

I didn't know this, maybe everyone else did...

The Matlab statement:

 clear all

clears breakpoints. So if you have "clear all" in your script, any breakpoints after that line will be cleared before script execution reaches them, so Matlab won't break. Getting rid of this line was the solution in my case, but it may not work in every case.

[opinion] this functionality is DUMB and should be fixed. [/opinion]

[Robert]

Quote from original post:

> I don't have any "clear all" statements

The solution was to stick with R2010b, or to wait for R2011b.

But agreed, it would be nice to have a "clear all but breakpoints" command.

[Jan]

Has this problem been fixed in the service pack2 for 2010b?