Thanks Olly for following-up! I have applied your patch and I can confirm that it fixes the issue. This is the diff between the two versions:
$ diff -u 1.csv 2.csv --- 1.csv 2016-01-13 19:05:07.497126268 -0300 +++ 2.csv 2016-01-13 19:06:01.069125169 -0300 @@ -3,20 +3,20 @@ ,,,,,,, ,,,,,,, ,"Dates 1","Integer Formatting","Float Formatting","Percent Formatting","Currency Formatting","Mixed Formatting", -,"39479","19","11136","0.510000305175781","15.1136","0.130000076293945", +,"39479","19","11136","0.51","15.1136","0.13", ,"39814","19","11992.1062271062","0.0023","15.6330005770485","12150", ,"40360","20","12150.666","0.00033","15.7355479452055","12150", ,"40634","20","-12092","0.0042","15.7297479452055","12150", -,"40909","20","11540","0.120000076293945","15.6745479452055","341.5603125", -,"41275","20","10741.9178082192","0.880000610351563","15.5947397260274","150", +,"40909","20","11540","0.12","15.6745479452055","341.56", +,"41275","20","10741.9178082192","0.88","15.5947397260274","150", ,,,,,,, ,"CASE 2",,"Hyperlink 1",,"HORIZONTALLY MERGED CELLS", ,,,,,, ,"Dates 2","Background Patterns 1","Background Patterns 2","Borders 1","Borders 2","Borders 3" -,"39479","23.20001953125","7.5","13125","19","15480" +,"39479","23.2","7.5","13125","19","15480" ,"39814","24.6575342465753","7.5","13125","19","15480" ,"40360","24.6575342465753","7.5","5062.5","18.3324350238071","15480" -,"40634","24.6575342465753","12","5062.5","15.450009765625","11610" +,"40634","24.6575342465753","12","5062.5","15.45","11610" ,"40909","24.6575342465753","12","5062.5","14.070609303653","11610" ,"41275","24.6575342465753","12.5","5062.5","14.0241509703196","11610" ,,,,,,, On 04/01/16 20:56, Olly Betts wrote: > Control: reopen -1 > Control: found -1 1:0.94.3~git20151007.fd634c2+dfsg-3 > > On Wed, Dec 23, 2015 at 11:15:04PM +0000, Debian Bug Tracking System wrote: >> It seems that this fix was already added upstream at some point before >> the migration to git (so I can't verify when did it happen). >> >> Since I have just discarded the Debian fork of catdoc, this issue should >> be fixed now, please verify. > > I can reproduce this, but my patch is actually reversed (which is why it > looks like it was applied upstream) - sorry about that, and the resulting > confusion. > >> I could not duplicate your bug anyway, maybe it was specific to Debian's >> 0.94.3, and that's not available in the archive any more. > > It seems the XL format has multiple ways to represent floating point > numbers, so just creating a document with the value 12.34 with > libreoffice isn't enough to make a reproducer. > > I've attached an existing file which shows the problem. It's a sample > from wxPython, so should be under a DFSG licence. > > For example, E8 should be 0.51, but xls2csv gives 0.510000305175781; E12 > should be 0.12, but xls2csv gives 0.120000076293945. With my patch > reverse-applied, I get the expected values. > > Cheers, > Olly > -- Martín Ferrari (Tincho)
