You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
test_fix.pro checks for the results of FIX() for negative inputs and unsigned returned types. The IDL docs (http://www.exelisvis.com/docs/FIX.html) don't reference this situation explicitly, but it seems like an inherently undefined operation and as such architecture/implementation specific. And while gdl appears to match IDL on x86_64, e.g.:
The test comments by Alain specifically refer to problems with negative inputs reported by a "Nicolas". Everybody could remove the test for negative values, but I would prefer Alain reviews this. Perhaps we should make the test only for intel architectures, or simply report the errors but let the test pass.
FIX has now /IMPLIED_PRINT (related to #48).
The various integer conversions have to be re-checked, as there are subtle differences (try print, long(!values.f_infinity) )
As reported by @opoplawski on SF.net (https://sourceforge.net/p/gnudatalanguage/bugs/622/):
test_fix.pro checks for the results of FIX() for negative inputs and unsigned returned types. The IDL docs (http://www.exelisvis.com/docs/FIX.html) don't reference this situation explicitly, but it seems like an inherently undefined operation and as such architecture/implementation specific. And while gdl appears to match IDL on x86_64, e.g.:
on arm this is returning:
I have no access to IDL on non-x86 architectures so I can't compare. But my suggestion would be to not test this situation.
The text was updated successfully, but these errors were encountered: