fib_sr: unittests for fib_sr fail sporadically

[miri64]

$ CFLAGS="-gdwarf-3 -g" make -C tests/unittests/ tests-fib_sr test
make: Entering directory `/home/martine/Repositories/RIOT-OS/RIOT/tests/unittests'
touch /home/martine/Repositories/RIOT-OS/RIOT/tests/unittests/main.c
Building application "unittests" for "native" with MCU "native".

"make" -C /home/martine/Repositories/RIOT-OS/RIOT/boards/native
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/boards/native/drivers
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/core
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/cpu/native
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/cpu/native/periph
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/drivers
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/sys
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/sys/embunit
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/sys/net/network_layer/fib
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/sys/universal_address
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/sys/xtimer
"make" -C /home/martine/Repositories/RIOT-OS/RIOT/tests/unittests/tests-fib_sr
   text    data     bss     dec     hex filename
  65437     416   52196  118049   1cd21 /home/martine/Repositories/RIOT-OS/RIOT/tests/unittests/bin/native/unittests.elf
RIOT native interrupts/signals initialized.
LED_GREEN_OFF
LED_RED_ON
RIOT native board initialized.
RIOT native hardware initialization complete.

main(): This is RIOT! (Version: 2016.03-devel-25-gab9f0f-beutlin-ipv6_hdr/fix/2015.12/ipv6_addr_dep)
......
fib_sr_tests.test_fib_sr_06_create_sr_with_hops (tests/unittests/tests-fib_sr/tests-fib_sr.c 264) exp 0 was -2

fib_sr_tests.test_fib_sr_06_create_sr_with_hops (tests/unittests/tests-fib_sr/tests-fib_sr.c 290) exp 0 was -2
......
run 12 failures 2
XXX: lpm_set(): LPM_POWERDOWN not implemented

Note, that sometimes they'll pass regardless.

[OlegHahm]

Tried about 35 times on SAMR21-xpro. Could not reproduce.

[miri64]

On what branch did you try (mine was on release branch (or as in the example above a branch forked from the release branch))

[miri64]

Oh and it was on native. Never tested it for any other platform actually, yet.

[OlegHahm]

IMO the check here: https://github.com/RIOT-OS/RIOT/blob/master/sys/net/network_layer/fib/fib.c#L911 doesn't make much sense and should be removed. @BytesGalore, @cgundogan?

This function (fib_sr_entry_add()) adds a new entry, right? Why should we check its lifetime here? And who should initialize its lifetime?

[OlegHahm]

@BytesGalore, please check if #4519 seems valid to you.

[BytesGalore]

@authmillenon @OlegHahm wow never bumped this problem I will have a look.
@OlegHahm fib_sr_entry_add() adds an entry behind a given one.
The function can be used at any time by the routing protocol, so it can also be called when the source route is expired.
The function enables to update existing source routes without the need to be recreate everything for small changes.
Usually, as a routing protocol, you would use fib_sr_entry_append() for putting a new source route to the FIB :).

[miri64]

Well they are there and #4519 does not fix this.

[miri64]

But it seems to be related to the fib_sr_check_lifetime() function (which in some cases returns -ENOENT).

[miri64]

Its due to the sr_lifetime member of the given sr_fib being somewhere around 1000000. Does this ring a bell @BytesGalore? I can't really set a watchpoint on this, since it seems to be timing related (setting the watchpoint causes the tests always to succeed).

[OlegHahm]

#4519 would fix it, but is apparently the wrong solution. ;)

[cgundogan]

got automatically closed by merging #4524. However, as I understand it, this issue still persists. Reopened.

[kaspar030]

Fixed by #4585.

[miri64]

Needs to be merged first.

[cgundogan]

Can be closed as #4585 was merged into master