The original bug is reported as https://bugs.gentoo.org/628242
I think test fails due to slight precision variation caused by FPU being 80-bit internally. If we add the following to CFLAGS, then: -mfpmath=387 (default on i686): test fails -mfpmath=sse (default on x86_64): test passes Quick reproduce howto: $ ./configure CFLAGS="-O2 -m32 -mfpmath=387" LDFLAGS=-m32 $ make testfixture $ ./testfixture test/date.test date-2.2c-1... ! date-2.2c-1 expected: [06:28:00.001] ! date-2.2c-1 got: [06:28:00.000] date-2.2c-2... Ok date-2.2c-3... Ok date-2.2c-4... ! date-2.2c-4 expected: [06:28:00.004] ! date-2.2c-4 got: [06:28:00.003] date-2.2c-5... Ok date-2.2c-6... Ok date-2.2c-7... ! date-2.2c-7 expected: [06:28:00.007] ! date-2.2c-7 got: [06:28:00.006] date-2.2c-8... ! date-2.2c-8 expected: [06:28:00.008] ! date-2.2c-8 got: [06:28:00.007] date-2.2c-9... Ok date-2.2c-10... ! date-2.2c-10 expected: [06:28:00.010] ! date-2.2c-10 got: [06:28:00.009] date-2.2c-11... ! date-2.2c-11 expected: [06:28:00.011] ! date-2.2c-11 got: [06:28:00.010] date-2.2c-12... Ok date-2.2c-13... Ok date-2.2c-14... ! date-2.2c-14 expected: [06:28:00.014] ! date-2.2c-14 got: [06:28:00.013] SQLite 2017-09-20 18:47:51 8e89ade69e1499a061b430c9dac2a13a2afd21281737207720f9ccbc7b41e6bf 464 errors out of 1595 tests on sf Linux 32-bit little-endian Note the drift here. Worth disabling tests on i?86? -- Sergei
pgpVw78JO_jW1.pgp
Description: Цифровая подпись OpenPGP
_______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
