GNOME Bugzilla – Bug 390498
test-numeric fails weirdly on windows
Last modified: 2018-06-29 21:20:04 UTC
See the attached file for the make output. 2005 tests appear to fail on win32. The code works fine on Mac and Windows.
Created attachment 79011 [details] "make test" output The attached make output shows all 2005 test failures from test-numeric on win32. Strangely, some of the math looks correct, but the code is still behaving strangely. Again, this only seems to fail on win32. The same code works on Linux and Mac. I know Andreas saw this, too, as he pointed it out to me first.
Created attachment 79198 [details] Output of the qofmath128 test routiones on Linux and on Windows This is the output of the lib/libqof/qof/qofmath128.c test functions. First part of file: Linux output; second part of file: Windows output. Note that the initial differences start already in the third line, 4294967297 * 4294967296, where the second to last argument shows the lower 64-bit integer in hexadecimal, and it is zero on windows but nonzero on Linux. I can't tell which one is correct, though.
Created attachment 79206 [details] output as of r15303 The problem was the printf(). I've fixed this in r15303, as you can see from the attached test output from win32. This new output looks correct and matches what we see on Linux. So I'm fairly confident that the qofmath routines are working as they should on win32.
Make check works now everywhere except src/bin. At least the following revisions are involved: r15472, r15473, r15474, r15480, r15482, r15483, r15484, r15493. Verification would be nice :)
GnuCash bug tracking has moved to a new Bugzilla host. This bug has been copied to https://bugs.gnucash.org/show_bug.cgi?id=390498. Please update any external references or bookmarks.