problems with gnumerics tests in a high precision version, where do the 'target values' originate from?

hello @all,

i tried to come over the problem with 'too long and thus not accurate values' stored by gnumeric for 'doubles' in the test files, by restricting values to 16 significant decimal digits for write to *.ods files. not a final solution, but a step.

now i can run the tests, and with few corrections where still 'odd' values came into the file they hold (for the moment just working with the first, statfuns.xls, mean i had 77,4000~0001 in B14)

when i try to request improved precision in the long double version (to 1E-14) i get some fails, 'KURT', 'SKEW' and 'TINV', not in total but one of the three samples for each.

may i think the 'compare values' are just taken from Excel or other less precise calculations and change them? tried that and with the values calculated with 'long'-version gnumeric standard still is! in the 1E-13 limit.

or do the compare values have other meaning, either 'this is correct' (and gnumeric long double calculates garbage), or 'we want / need excel compatibility and thus need these results'?

additional questions, i can change the output to ods files by patching openoffice-write.c, and can inspect the created *.ods files by opening with an archive manager, but both not for *.xls. As quite many test samples are *.xls: where in the code is defined how many digits should be exported to excel files, and doe's anybody know a unzipper for xls files?

best regards, TIA for any help,


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]