[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#671550: [Pkg-javascript-devel] Bug#671550: nodejs: building on kfreebsd-*

On 16/06/2012 15:03, Steven Chamberlain wrote:
> Hi,
> On 15/06/12 21:55, Jérémy Lal wrote:
>> Did you rebuild libv8 after i applied your patches (i modified them) ?
>> All tests passed ?
> I've pulled the latest libv8 git and rebuilt it;  all tests passed.
> Here is a log of that:
> http://pyro.eu.org/f/3gP6A9d4RAKl5U_X0rFHIw.txt
> The test harness (Python) seems to hang _some_ of the time;  I can't
> always reproduce it, but this is a known issue with Python threading (or
> glibc) on kfreebsd-*, seen occasionally on the buildds when building
> some other packages.  If I left it long enough (there's a 180-minute
> timeout?) maybe it would be able to recover and finish building.

180 seconds :)

>> Also could you paste the build log of nodejs ? I'd like to understand why
>> nine tests are failing.
> I don't have a log of that.  So I did a rebuild from scratch using the
> libv8 I just compiled, latest nodejs package from git, and the most
> recent set of patches I sent you.  This time there were only 5 test
> failures:
> http://pyro.eu.org/f/StBvXuZ7Sn6urlCwb13jSg.txt

Notice the

that are on that build log, and that are not on the other ones.
I guess it's partly building c-ares, or using headers from c-ares,
which is wrong since c-ares is a build-dependency.

> This failure is new in 0.6.19~dfsg1-3:
> * simple/test-http-full-response
>> assert.js:104
>>   throw new assert.AssertionError({
>>         ^
>> AssertionError: 3 == 0
> There seems to be a timeout when the test runs this:
> ab -c 1 -n 10
> But it works okay if I run this 10 (or even 100) times instead:
> ab -c 1 -n 1
> The other tests were already failing in 0.6.16~dfsg1-2:
> * test-fs-utimes -- maybe not supported on ZFS?
> * test-fs-watch -- fails with ENOSYS = not supported
> * test-setproctitle -- expected to fail on kfreebsd-* (it is not
> implemented yet in libbsd)
> * test-http-upgrade-server:
> That looked like a response is being served with "Transfer-Encoding:
> chunked" which the test case doesn't understand.  So if I change the
> request type to HTTP/1.0 on test/simple/test-http-upgrade-server.js:155
> it passes.

Ok, i propose to let nodejs build fail on those tests, just to see what
failures comes out on kfreebsd buildd server(s). Then i'll fix or forward
them upstream.

Thank you.


Reply to: