GNOME Bugzilla – Bug 344950
Mis-import of marker type from xls
Last modified: 2008-04-11 23:01:55 UTC
Steps to reproduce the problem: - open the file here in gnumeric: http://www-lgge.obs.ujf-grenoble.fr/~parrenin/download/bugs/IGFB23-1.xls => the calculations on sheet1 are OK - now go to sheet2, and change value of cell E1 from 0 to 1 => a lot of cells in sheet1 now equal 0 - change back the value of E1 from 1 to 0 => the values of the cells in sheet1 do not go back to their initial value
Just loading the sheet and hitting F9 to force a recalc makes, e.g., cell C21 turn to #REF! The formula there is... =if(A21<$I$15+1,vlookup($I$15+A21,'Paste raw data here'!$E$1:index('Paste raw data here'!$I$1:$I$999,'Results sheet'!$I$14),2,TRUE),0) ...which suggests a vlookup or index failure. A21 abd I15 are both integers.
Interestingly, that expression does not parse. I get "Constructed ranges use simple references". Loading also produces... Hmm, ptgAttr of type 0 ?? I've seen a case where an instance of this with flag A and another with flag 3 bracket a 1x1 array formula. please send us this file. Flags = 0x3
The URL for the spreadsheet showing this problem is no longer working. Can the original reporter (or anyone else who has a copy) repost it or attach it to this bug?
You can find the file here: http://parrenin.frederic.free.fr/download/bugzilla/IGFB23-1.xls
I am not seeing the issues from comment #1 with svn trunk. Nor are there substantial differences when toggling E1 However, we are mis-importing the marker shapes in the chart, and stacking them in the wrong order.
Created attachment 99193 [details] [review] fixes the marker issue The charts have default marker == none. Most probably we need to use auto_shape instead of the current test (marker != none). The patch does not seem to break import of chart-tests-excel.xls
Fixed both the marker shapes and stack order.
Sounds like we can close this. Please test with 1.7.90
Hmm, seems I forgot to commit part of the fix. Commited now, test should be done with 1.7.91.
Can we close this ?
As far as I am concerned, yes.