Closed malvira closed 11 years ago
econotag is bad b/c link to the compiler isn't good (was at adamdunkles.github.com)
following patch will fix it:
https://github.com/malvira/contiki/commit/ccccd461b57d72a41f06a8984ce06519c42fd5c9
$ [ $MAKE_TARGETS = cooja ] && java -version && ant -q -f tools/cooja/build.xml jar || true
/home/travis/build.sh: line 144: [: =: unary operator expected
might be the cause
download link seems fine...
SDCC builds broke at #173. I've got a fix already in place and will open a pull in due course (when the global IPv6 switch discussion has reached a conclusion).
That discussion consluded I believe...
http://sourceforge.net/mailarchive/message.php?msg_id=30795889
Is there a PR that reverts Adam's goof?
I modified the regression test makefile to print out the faillog so we can see why travis compiles fails. Example here:
https://www.travis-ci.org/malvira/contiki/jobs/7264845
this has the change:
https://github.com/malvira/contiki/commit/abf2fa90bcf6ff8f7065acd830d268740659148d
hello-world/ avr-raven: FAIL ಠ.ಠ
tcpip.c:(.text.tcpip_ipv6_output+0x238): undefined reference to `uip_ext_len'
contiki-avr-raven.a(tcpip.o): In function `process_thread_tcpip_process':
tcpip.c:(.text.process_thread_tcpip_process+0x6e): undefined reference to `rpl_init'
tcpip.c:(.text.process_thread_tcpip_process+0x19c): undefined reference to `uip_ds6_timer_periodic'
tcpip.c:(.text.process_thread_tcpip_process+0x19e): undefined reference to `uip_ds6_timer_periodic'
tcpip.c:(.text.process_thread_tcpip_process+0x1b2): undefined reference to `uip_ds6_periodic'
collect2: ld returned 1 exit status
can you point me at a commit to cherry-pick? I'm making a big PR that fixes everything related to the travis build problem. (it includes flipping the master switch back).
On Fri, May 17, 2013 at 5:57 PM, George Oikonomou notifications@github.comwrote:
SDCC builds broke at #173https://github.com/contiki-os/contiki/issues/173. I've got a fix already in place and will open a pull in due course (when the global IPv6 switch discussion has reached a conclusion).
— Reply to this email directly or view it on GitHubhttps://github.com/contiki-os/contiki/issues/226#issuecomment-18088169 .
I'll do the master switch and the SDCC stuff (I was in fact writing the pull message right now)
On 18 May 2013, at 00:02, Mariano Alvira wrote:
can you point me at a commit to cherry-pick? I'm making a big PR that fixes everything related to the travis build problem. (it includes flipping the master switch back).
On Fri, May 17, 2013 at 5:57 PM, George Oikonomou notifications@github.comwrote:
SDCC builds broke at #173https://github.com/contiki-os/contiki/issues/173. I've got a fix already in place and will open a pull in due course (when the global IPv6 switch discussion has reached a conclusion).
— Reply to this email directly or view it on GitHubhttps://github.com/contiki-os/contiki/issues/226#issuecomment-18088169 .
— Reply to this email directly or view it on GitHub.
perfect. I think i'm pretty close to fixing travis. Now that I've got my first green check I want to get them all :)
On Fri, May 17, 2013 at 7:04 PM, George Oikonomou notifications@github.comwrote:
I'll do the master switch and the SDCC stuff (I was in fact writing the pull message right now)
On 18 May 2013, at 00:02, Mariano Alvira wrote:
can you point me at a commit to cherry-pick? I'm making a big PR that fixes everything related to the travis build problem. (it includes flipping the master switch back).
On Fri, May 17, 2013 at 5:57 PM, George Oikonomou notifications@github.comwrote:
SDCC builds broke at #173< https://github.com/contiki-os/contiki/issues/173>. I've got a fix already in place and will open a pull in due course (when the global IPv6 switch discussion has reached a conclusion).
— Reply to this email directly or view it on GitHub< https://github.com/contiki-os/contiki/issues/226#issuecomment-18088169> .
— Reply to this email directly or view it on GitHub.
— Reply to this email directly or view it on GitHubhttps://github.com/contiki-os/contiki/issues/226#issuecomment-18090298 .
looks like twitter was only partially removed and broke a bunch of stuff. Seems like most of the cooja tests are working again now.
Ok, I think we are down to a handful of cooja errors: I'm going to need help with this. I have the tests printing out the logs now when cooja fails. This is essential to debugging the output of travis. Here is problem number 1:
make: Entering directory `/home/travis/build/malvira/contiki/regression-tests/12-rpl'
Running test 01-rpl-up-route ... FAIL ಠ_ಠ
Random seed: 123456
did not add sink as it was already there
FAIL ಠ_ಠ
OK
Running test 02-rpl-root-reboot ... FAIL ಠ_ಠ
Random seed: 123456
did not add sink as it was already there
FAIL ಠ_ಠ
OK
Running test 03-rpl-28-hours ... FAIL ಠ_ಠ
Random seed: 123456
did not add sink as it was already there
FAIL ಠ_ಠ
OK
Running test 04-rpl-large-network ... FAIL ಠ_ಠ
Random seed: 123456
did not add sink as it was already there
FAIL ಠ_ಠ
OK
make: Leaving directory `/home/travis/build/malvira/contiki/regression-tests/12-rpl'
please help :)
BTW my travis for contiki is here: https://www.travis-ci.org/malvira/contiki
Some more:
Running test 01-cooja-collect ... FAIL ಠ_ಠ
836000 34 Starting 'Test collect process'
1483000 25 Contiki 2.6 started. Node id is set to 25.
1483000 25 Rime started with address 25.0
...
3425000 47 Starting 'Test collect process'
FAIL ಠ_ಠ
Running test 10-cooja-mesh ... Aborted
FAIL ಠ_ಠ
looks like twitter was only partially removed and broke a bunch of stuff. Seems like most of the cooja tests are working again now.
I merged that twitter removal (or rather rename so it might be added back under the new name?)
Sorry for breaking something - but as soon as Travis usualy works something like that (hopefully) won't go unnoticed anymore.
Yeah, it seems like Travis broke/got ignored and then it was impossible to know if we were breaking stuff. Looks last we were :) So it goes.
Its also possible that those remaining cooja tests are legitimate failures of f the simulation...
-Mar. On May 18, 2013 3:02 AM, "Oliver Schmidt" notifications@github.com wrote:
looks like twitter was only partially removed and broke a bunch of stuff. Seems like most of the cooja tests are working again now.
I merged that twitter removal (or rather rename so it might be added back under the new name?)
Sorry for braking something - but as soon as Travis usualy works something like that (hopefully) won't go unnoticed anymore.
— Reply to this email directly or view it on GitHubhttps://github.com/contiki-os/contiki/issues/226#issuecomment-18096276 .
It looks to me that java in ubuntu went from 1.7.0_09 to 1.7.0_15 and that is causing cooja to fail now. I re-ran an old job that passed and now it's failing. This happenend about 2 months ago which also coincided with the econotag toolchain link failing (and travis has been broken since).
I say if you have all of the compile issues in travis fixed and you have at least some of the cooja tests running, then simply disable the tests that are broken for now and merge it in. No need to be "perfect" if it's going to take too long and you aren't getting traction.
I think the next one will be all green (once it runs). I found the java default that changed. Ubuntu disables "shared" by default now and cooja requires that.
The only thing that's not the "right" fix is maybe the master switch. But I'm happy to merge what we have on that now (since it will fix travis).
On Sat, May 18, 2013 at 2:03 PM, Robert Quattlebaum < notifications@github.com> wrote:
I say if you have all of the compile issues in travis fixed and you have at least some of the cooja tests running, then simply disable the tests that are broken for now and merge it in. No need to be "perfect" if it's going to take too long and you aren't getting traction.
— Reply to this email directly or view it on GitHubhttps://github.com/contiki-os/contiki/issues/226#issuecomment-18105405 .
Ok everything I just merged should almost make Travis good. The only fixes left to complete are the "rimestats" related fixes. George has the best work on that so far here:
https://github.com/g-oikonomou/contiki/tree/bug-fixes/rimestats
If we get that in, then the board should go all green :)
On Sat, May 18, 2013 at 2:07 PM, Mariano Alvira mar@devl.org wrote:
I think the next one will be all green (once it runs). I found the java default that changed. Ubuntu disables "shared" by default now and cooja requires that.
The only thing that's not the "right" fix is maybe the master switch. But I'm happy to merge what we have on that now (since it will fix travis).
On Sat, May 18, 2013 at 2:03 PM, Robert Quattlebaum < notifications@github.com> wrote:
I say if you have all of the compile issues in travis fixed and you have at least some of the cooja tests running, then simply disable the tests that are broken for now and merge it in. No need to be "perfect" if it's going to take too long and you aren't getting traction.
— Reply to this email directly or view it on GitHubhttps://github.com/contiki-os/contiki/issues/226#issuecomment-18105405 .
Ok it's fixed now :) If you rebase your pull requests then you can test them against a good travis and contiki without build errors.
You, sir, are awesome. Good job!
Travis has multiple failures.