Forgot to mention: Ubuntu 14.04, 32 bit

Hi @andis,

Err:522 as you can see in the status bar means a "circular reference", it happens when a cell is included in the calculation of their formula (directly or indirectly)

There is an option to allow calculate with circular references:
Menu/Tools/Options/LibreOffice calc/Calculate - Iterative references.
that avoid the errors.

Closed as not a bug, please if you are not agree, reopen it.

Thank you for advice in solving this problem! 

I find mistake in the initial report - it applies to 5.1.4.2!

Iterative references is not really solution, because it is possible to get more compatibility and calculation result related problems in future by activating the Iterative references function.

Formulas in attached file works well in Libreoffice 5.0.x, Openoffice.org 4.1.2 and Microsoft Excel and Iterative references are not activated in Libreoffice 5.0.x and Openoffice.org 4.1.2. Therefore to my understanding, keeping in mind hundreds or thousands of files, which now shows error instead of result in 5.1.x, it is serious regression in basic calc functions.

I reopened the report, because I messed up versions of Libeoffice at the beginning.

I did my test with:
Win10x64
Version: 5.1.4.2 (x64)
Build ID: f99d75f39f1c57ebdd7ffc5f42867c12031db97a
CPU Threads: 1; OS Version: Windows 6.19; UI Render: GL

The matter it's the same with:
Version: 5.0.6.2 (x64)
Build ID: b3fbfa99158a1030fb79f0ba72b6851afc3c7895-GL
Version: 5.3.0.0.alpha0+
Build ID: 757f221bceb74ccc2af8a9d4de149076280f29fb
CPU Threads: 4; OS Version: Windows 6.19; UI Render: GL; 
TinderBox: Win-x86@39, Branch:master, Time: 2016-06-23_23:54:55

Err:522 it's there, and it is a 'circular reference' error, if 'Iterative references' it's not enable.

The only bug could be if there wasn't circular references, but C4 calls B4, while B4 has COUNTA($B4:$T4) where C4 is part of the range.

So for not a bug, you can ask for help in ask.libreoffice.org or in mailing list http://nabble.documentfoundation.org/Users-f1639498.html

Thank you for quick response!

This issue really is not a big problem. However, circular reference this time is recognized by mistake. COUNTA calls to number of cells with no relation to values and it is strange change in compare to previous versions and to other office suites. Therefore to me it is regression.

Note this in the help for COUNTA, values matter in the cells:
"If an argument is an array or reference, empty cells within the array or reference are ignored."

I think that makes it clear it's intended behavior, and not a bug, since the circular reference exists because of the definition. What do you think Andis?

The behavior was changed in v4.1.0.4, by the way.

Hello!

This is manageable, however confusing for those used to Excel or Openoffice.org Scalc. Probably, it is worth to understand, why this "feature" is introduced and may be there is easy way to get rid of it to keep compliance with other spreadsheet applications.
Considering limited resources instead of this bug I would definitely vote for increase of performance of Scalc, which is becoming more and more evident problem with every release.

> 
> I think that makes it clear it's intended behavior, and not a bug, since the
> circular reference exists because of the definition. What do you think Andis?

Hi! I checked Excel's description for COUNTA as well, and based on that it seems you're right, the circular reference can be eliminated based on this criteria: "The COUNTA function counts cells containing any type of information, including error values and empty text ("")".

So regardless of what the field with formula in the range evaluates to, it can be counted in COUNTA (I don't know what actually happens in Excel, this is just my deduction).

For reference and comparison:
https://support.office.com/en-us/article/COUNTA-function-7dc98875-d5c1-46f1-9a82-53f3219e2509
https://help.libreoffice.org/Calc/Statistical_Functions_Part_One#COUNTA

Adding needAdvice keyword to get input on this from developers.

Now that we discussed COUNTA, is anything wrong with COUNT?

Created attachment 132327 [details]
The original file saved as a MS Excel 2003 file

Created attachment 132328 [details]
The original file saved as a MS Excel 2007-2013 file

Can anyone test this files using MS Excel?
Does it warn about "circular references"?

I have set the bug's status to 'NEEDINFO'. Please change it back to 'UNCONFIRMED' once the requested information is provided.

(In reply to Carlos from comment #11)
> 
> I have set the bug's status to 'NEEDINFO'. Please change it back to
> 'UNCONFIRMED' once the requested information is provided.

In both cases Excel 2007 opens file with warning on circular references.

Hello!

There is no inquiry for additional information. It was not decided, if this is bug or improvement. There is more or less easy workaround or switching to OpenOffice.org, therefore to me this report can be closed.

Regards

Seems like this was fixed by the same commits as bug 118161, in 6.2 / 6.1 / 6.0.5.