Debian Patches
Status for rauc/1.8-2
| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| disable-network-tests.patch | Disable tests that fail in debian build chroot | Arnaud Rebillout <arnaud.rebillout@collabora.com> | not-needed | vendor, Debian | 2018-12-11 | |
| install-wrapper-to-pkgdatadir.patch | Install dbus wrapper script to /usr/share/rauc instead of /usr/libexec According to the FHS /usr/libexec is for binaries only, so /usr/share/rauc seems to be better suited for a shell script. |
Uwe Kleine-König <uwe@kleine-koenig.org> | no | 2019-06-13 | ||
| test-stats-make-tests-involving-floating-point-calcu.patch | test/stats: make tests involving floating point calculations more lax On i387 floating point calculations can be quite surprising. The motivation for this change came from the observation that compiling and testing rauc-1.8 on the i386 Debian build machine had a failure on g_assert_cmpfloat(r_stats_get_avg(stats), ==, 62.54615384615385); . In that case r_stats_get_avg(stats) (when stored in a double typed variable) returned 0x1.f45e85e85e85fp+5 which is 0x1f45e85e85e85f00 / 2**55. However when stored in a long double typed variable it has the value 0x1.f45e85e85e85e85p+5 This value is 0x1f45e85e85e85e85 / 2**55 and so it has a significand that cannot be stored in the 53 bits that are allocated for double typed variables. 62.54615384615385 is a plain double and so evaluates to 0x1.f45e85e85e85fp+5. This makes the comparison implemented in g_assert_cmpfloat() (which uses long double precision) fail. To fix that, use g_assert_cmpfloat_with_epsilon() and also adapt the other tests that involve non-trivial calculations to this approach. |
Uwe Kleine-König <u.kleine-koenig@pengutronix.de> | yes | 2022-10-26 |