fix points write panic

[oiooj]

fix index out of range risk, log here:

[wal] 2015/10/28 11:39:55 write to index of partition 1 took 3.516488ms
[wal] 2015/10/28 11:40:05 Flush due to idle. Flushing 8 series with 8 points and 424 bytes from partition 1
[wal] 2015/10/28 11:40:05 write to index of partition 1 took 3.126415ms
[wal] 2015/10/28 11:40:15 Flush due to idle. Flushing 8 series with 8 points and 424 bytes from partition 1
[wal] 2015/10/28 11:40:15 write to index of partition 1 took 1.716188ms
[http] 2015/10/28 11:40:25 221.221.221.98 - - [28/Oct/2015:11:40:25 +0800] POST /write HTTP/1.1 400 57 - Mozilla/5.0 (Macintosh; Intel Mac 
OS X 10_11_1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/46.0.2490.71 Safari/537.36 9facaff5-7d25-11e5-8001-000000000000 2.522487ms
[wal] 2015/10/28 11:40:25 Flush due to idle. Flushing 8 series with 8 points and 424 bytes from partition 1
[wal] 2015/10/28 11:40:25 write to index of partition 1 took 12.511171ms
[wal] 2015/10/28 11:40:35 Flush due to idle. Flushing 8 series with 8 points and 442 bytes from partition 1
[wal] 2015/10/28 11:40:35 write to index of partition 1 took 7.997816ms
[http] 2015/10/28 11:40:38 221.221.221.98 - - [28/Oct/2015:11:40:38 +0800] POST /write HTTP/1.1 400 45 - Mozilla/5.0 (Macintosh; Intel Mac 
OS X 10_11_1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/46.0.2490.71 Safari/537.36 a776548d-7d25-11e5-8002-000000000000 19.391701ms
[wal] 2015/10/28 11:40:45 Flush due to idle. Flushing 8 series with 8 points and 442 bytes from partition 1
[wal] 2015/10/28 11:40:45 write to index of partition 1 took 1.890025ms
[http] 2015/10/28 11:40:50 221.221.221.98 - - [28/Oct/2015:11:40:50 +0800] POST /write HTTP/1.1 200 23 - Mozilla/5.0 (Macintosh; Intel Mac 
OS X 10_11_1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/46.0.2490.71 Safari/537.36 af16a5b9-7d25-11e5-8003-000000000000 1.222898ms [pan
ic:runtime error: index out of range]

[otoolep]

@oiooj -- can you please run the test suite locally before pushing up any further commits? That way you will get the quickest turnaround when debugging your code.

[oiooj]

@otoolep sure, I have ran the test file.

$ go test points_test.go 
ok      command-line-arguments  0.013s

$ go test handler_test.go 
ok      command-line-arguments  0.019s

It is right ?

[otoolep]

I don't know, depends if you are in the correct directory.

$ ls
balancer.go       circle-test.sh  cmd                       DOCKER.md  influxql                    Makefile  nightly.sh  README.md  snapshot  test-32bit-docker.sh  uuid
balancer_test.go  circle.yml      CONTRIBUTING.md           errors.go  influxvar.go                meta      package.sh  scripts    statik    tests
build-docker.sh   client          Dockerfile                etc        LICENSE                     models    pkg         services   stress    toml
CHANGELOG.md      cluster         Dockerfile_test_ubuntu32  importer   LICENSE_OF_DEPENDENCIES.md  monitor   QUERIES.md  shared     tcp       tsdb
$ cd models/
$ go test -run TestParsePointNoValue
PASS
ok      github.com/influxdb/influxdb/models     0.001s

will run the test that is failing on Circle.

[oiooj]

No problem here, I don't know what's going on. Always failed on Circle.

Sosara@MBA:~/golang/src/github.com/oiooj/influxdb/models$ go test -run TestParsePointNoValue
PASS
ok      github.com/oiooj/influxdb/models        0.010s

and the circle-test.sh

...
--- PASS: TestWAL_Cursor_Reverse (0.00s)
PASS
ok      github.com/influxdb/influxdb/tsdb/engine/wal    0.949s
?       github.com/influxdb/influxdb/tsdb/internal  [no test files]

? github.com/influxdb/influxdb/uuid [no test files] bash circle-test.sh returned exit code 1

[oiooj]

@otoolep fixed , Thanks

[otoolep]

Thanks @oiooj.

Can you please rebase and squash the commits, with a single commit?

[otoolep]

Can you also add a test case for this? We really need test cases for every change.

[oiooj]

rebased the commits , I reviewed handler_test.go source, This is a post and write data request, It's difficult to add a test case. Thanks @otoolep

[otoolep]

OK, I see what you are doing. Basically, if it's not JSON by index (len-1) (and not a valid ASCII character either), it's not JSON at all. Is that what you're getting at?

[otoolep]

I believe I follow what this patch fixes, but I'd like a second set of eyes.

@corylanou ?

[corylanou]

Yes this appears to make sense. We should add a test though so that we don't regress. Once a test is added, +1. Thanks!

[otoolep]

Closing in favour #4625