Bug#639636: [node-dev] Re: Bug#639636: nodejs: FTBFS/armel - testsuite fails
update : now building 0.6.8, test/simple/test-buffer is still failing on armel (armv4t), see https://buildd.debian.org/status/fetch.php?pkg=nodejsarch=armelver=0.6.8~dfsg1-1stamp=1327488576 Since the test used to succeed on armv7, i guess it needs an upstream v8 opinion. Jérémy. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#639636: [node-dev] Re: Bug#639636: nodejs: FTBFS/armel - testsuite fails
On 31/12/2011 02:47, Steven Chamberlain wrote: Hello, I wonder if this commit (introduced in 0.5.10) is what fixed the failure of test-buffer.js in later versions: https://github.com/joyent/node/commit/6a72e525202972fe5d5a89cc94f48071380c1aa7 I have no armel build system but would appreciate if someone could humour me by trying this :) I believe i already tried that commit on 0.4 branch, and it did not work. Jérémy. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#639636: [node-dev] Re: Bug#639636: nodejs: FTBFS/armel - testsuite fails
Hello, I wonder if this commit (introduced in 0.5.10) is what fixed the failure of test-buffer.js in later versions: https://github.com/joyent/node/commit/6a72e525202972fe5d5a89cc94f48071380c1aa7 I have no armel build system but would appreciate if someone could humour me by trying this :) I came across this when I noticed there are different offsets shown for the two buffers test-buffer.js compares: original buffer: 5684:252,5685:0,5686:98,5687:0,5688:101,5689:0,5690:114,5691:0,5692:0, 5693:252,5694:0,5695:98,5696:0,5697:101,5698:0,5699:114,5700:0,5701:4, offset:5693} copy-constructed buffer: 5684:252,5685:0,5686:98,5687:0,5688:101,5689:0,5690:114,5691:0,5692:0, 5693:252,5694:0,5695:98,5696:0,5697:101,5698:0,5699:114,5700:0,5701:4, offset:5685} So I wondered if this was due to only one byte being read or written -- stopping because of the null character -- that would leave the offset short of where it should be, which is incidentally 1 byte into a different copy of the string (and that is maybe what we were seeing). Thanks, Regards, -- Steven Chamberlain ste...@pyro.eu.org -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#639636: [node-dev] Re: Bug#639636: nodejs: FTBFS/armel - testsuite fails
On Mon, Dec 5, 2011 at 20:02, Jérémy Lal kapo...@melix.org wrote: On 05/12/2011 17:51, Jérémy Lal wrote: Thanks to the abel guest access i got, libv8 builds and pass tests on armel and armhf. I'm still having *one* test failing, if anyone has an idea about why : var f = new Buffer('über', 'ucs2'); assert.deepEqual(f, new Buffer([252, 0, 98, 0, 101, 0, 114, 0])); Giving this result : https://buildd.debian.org/status/fetch.php?pkg=nodejsarch=armelver=0.4.12-2stamp=1323100281file=log Couldn't be linked to some wrong endianness assumption ? It isn't, after doing some testing on abel, i observe this strange behavior : build/default/node /home/kapouer/nodejs/nodejs-0.4.12/test/simple/test-buffer.js never outputs errors, while python tools/test.py simple/test-buffer always does. I'm stuck. Jérémy. PS: wild cc-ing nodejs-dev, in case someone there already knows why :) the test mentionned above is the only one failing, and it's been there for a long time. It looks like a off-by-one error. To wit: {0:252,1:0,2:98,3:0,4:101,5:0,6:114,7:0 {0:0,1:98,2:0,3:101,4:0,5:114,6:0,7:120 It's quite possibly been fixed. Does it happen with v0.6.5? -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#639636: [node-dev] Re: Bug#639636: nodejs: FTBFS/armel - testsuite fails
On 05/12/2011 21:54, Ben Noordhuis wrote: On Mon, Dec 5, 2011 at 20:02, Jérémy Lal kapo...@melix.org wrote: On 05/12/2011 17:51, Jérémy Lal wrote: Thanks to the abel guest access i got, libv8 builds and pass tests on armel and armhf. I'm still having *one* test failing, if anyone has an idea about why : var f = new Buffer('über', 'ucs2'); assert.deepEqual(f, new Buffer([252, 0, 98, 0, 101, 0, 114, 0])); Giving this result : https://buildd.debian.org/status/fetch.php?pkg=nodejsarch=armelver=0.4.12-2stamp=1323100281file=log Couldn't be linked to some wrong endianness assumption ? It isn't, after doing some testing on abel, i observe this strange behavior : build/default/node /home/kapouer/nodejs/nodejs-0.4.12/test/simple/test-buffer.js never outputs errors, while python tools/test.py simple/test-buffer always does. I'm stuck. Jérémy. PS: wild cc-ing nodejs-dev, in case someone there already knows why :) the test mentionned above is the only one failing, and it's been there for a long time. It looks like a off-by-one error. To wit: {0:252,1:0,2:98,3:0,4:101,5:0,6:114,7:0 {0:0,1:98,2:0,3:101,4:0,5:114,6:0,7:120 It's quite possibly been fixed. Does it happen with v0.6.5? Thank you for your suggestion, Indeed, it doesn't. Now i'm searching why. It doesn't seem to be in src/node_buffer.cc : i just tried with a backported one and it doesn't change the failure, same goes for lib/buffer.js. Since it fails when launched trough test.py and not directly through command-line node invocation, does it point to something particular i could bisect on ? Note that i kept the same system libv8 version (3.5) for both 0.4.12 and 0.6.5 tests, it's unsupported but working quite well. Jérémy. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org