[ooo-build] [hardy] Speadsheet has problems in hardy

I'm adding a bit more info. Attached are two screenshots of frequency data sent to me. One was opened right after the other. Notice that the data is not the same. Both were opened by clicking on the download link in my email and choosing to open with OO.o. They opened in read only form. When the files was opened in gutsy this problem did not manifest. Only in Hardy and on two different machines.

Here is the second screenshot. Guess I can't add two to one post.

Okay, one last bit of info on this. It was just pointed out to me that values are not changing - they are being moved around. Still, not normal but maybe a bit of info that will help explain what is happening and how to fix it.

I can confirm the correct results in gutsy and and wrong results in hardy (OO 2.4.0). However, P3-P13 has wrong but different results as described (13, 16, 31, 24, 29, 50, 33, 30, 16, 2, 8).

Ralph, That is part of the problem. Each time it's opened the values have moved. Your results will be different from mine and different from your own the next time you open it.

For what it's worth, I did file this upstream.

http://www.openoffice.org/issues/show_bug.cgi?id=87954

This bug only seems to occur in the ubuntu build of OpenOffice.org 2.4.0. I've installed the Sun build of 2.4.0 (downloaded from openoffice.org) in Hardy beta and the values are displayed correctly. The results are also correct using OO.o 2.4.0 in windows xp.

I've installed the latest updates to hardy beta, which included updates to OO.o (ubuntu build), and the bug is still present.

Might be worth trying another array formula to see if this has similar issues or if its just the frequency function. I'll give this a go, but I don't have much experience with array formulas.

Mike, that is my experience too. I've used it in OSX (well it's NeoOffice there), windows and Ubuntu Gutsy and Hardy and only in Hardy do I see it. I don't know much about spreadsheets so it was just happenstance that I found this and recognized that something was wrong.

I've tried some other array formulas and there doesn't seem to be a problem with them.

From a quick look on openoffice.org (OOo), I think this bug is related to OOo issue 8946,

http://www.openoffice.org/issues/show_bug.cgi?id=8946

which had some changes to the frequency function at the end of last year and in February 2008 to allow for the values in the classes range not being in ascending order. The following comment was made by er on 6th Feb 2008:

"The results of FREQUENCY() were somewhat arbitrary because when shuffling the
sort array prior to invoking QuickSort the index order didn't follow. Fixed in

sc/source/core/tool/interpr3.cxx 1.20.30.8"

Strangely though, the "FREQUENCY_test.ods" file attached to OOo issue 8946 appears to work (with results in the correct order) in the OOo 2.4-ubuntu build but the OOo 2.4-sun build gives incorrect results when the values in the classes are not in ascending order.

I should also add that this bug of frequency formula results being displayed in a random order also occurs when using the example from the openoffice.org help file, see attached. As its only a small file, sometimes the results are displayed correctly but sometimes its in an apparently random order.

According to the OO.o site (http://www.openoffice.org/issues/show_bug.cgi?id=88477) this bug has been fixed and only affected ooo-build trunk. Has this fix made it's way to Hardy? I'm guessing no as I have an updated hardy and it is most definitely still a problem. This really undermines ones faith in this product. It's hard to do calculations and have confidence in them knowing such issues exist.

I've installed today's updates to hardy, which included openoffice.org 2.4.0-3ubuntu6, and the problem still exists.

This was fixed in ooo-build but was fixed too late for it to make it into Hardy. So it will be fixed in Ubuntu 8.10.

Fair enough if it can't be in the hardy release, but why not add it to the updates that invariably come a week or so after release? This isn't a small error. Anyone using frequencies will get wrong results. We pretty much have blind faith in the calculating power of computers. This will cause problems for people and should be fixed as soon as possible.

Ok, I will mark it as 8.04.1 and see if the RM will allow the patch in.

Accepted into -proposed, please test and give feedback here

Jim, can you confirm that the OpenOffice package in hardy-proposed fixes this problem as you saw it?

Steve, thanks for the info. I will check into it.

I can confirm that the spreadsheet attached by Jim now gives the correct results with the OO version in -proposed.

Jim, Ralph: Please see my blog post on how to do hardy-proposed testing to test for yourself: http://blog.omma.net/?p=11

If you can confirm the fix we can push out the update!

The hardy-proposed SRU was copied to intrepid some time ago.

Copied to hardy-updates.