The app runs out of file descriptors when using prepared statements

[mattn]

The app runs out of file descriptors when using prepared statements

[mattn]

I got following result.

$ ps aux | grep fuf
mattn    20516  0.0  1.1 807012  1468 pts/4    Sl+  09:17   0:00 ./fuf
$ while [ 1 ] ; do ls /proc/20516/fd | wc -l ; sleep 1; done
3
3
3
3
3
3
3
ls: cannot access /proc/20516/fd: No such file or directory
0
ls: cannot access /proc/20516/fd: No such file or directory
0
ls: cannot access /proc/20516/fd: No such file or directory
0

Are you using gccgo?

[ernestoalejo]

Using 8g, Fedora 17, 32 bits. Using the sqlite3.7.11-3.fc17 package.

My results look something like this (after introducing a time.Sleep of 20 Milliseconds to obtain more measurements):

3
3
3
3
5
5
11
16
21
27
32
37
43
49
.........more results here....
981
987
993
999
1004
1010
1016
1021     // here the program fails
0
0
0

Let me know if I can provide some more information to debug the issue.

[mattn]

Hmm, do you using latest version of go & sqlite3-devel?

I tried above on CentOS6 32bit.

[ernestoalejo]

Go installed from source:
devel +ca83aeaa55e0 Fri Feb 22 12:23:19 2013 +1100

SQLite3 is the latest version Fedora has (3.7.11), but not the current one (3.7.15.2).

I was about to try the same example using C, to see if it happens there too.

[ernestoalejo]

I've tested a similar example ported to C with the functions you use and no leak of descriptors.

Then I downgraded my Go version to 1, (hg update go1.0.3). The gist code produced no leaks there. After updating to the latest one (hg pull && hg update default) the loop stopped at iteration i=1021 with no file descriptors remaining. Same code in both cases.

Have you some distribution of Go around to test if it is a regression?

[mattn]

I've tested on tip.

CentOS6

2.6.32-042stab061.2 #1 SMP Fri Aug 24 09:07:21 MSK 2012 i686 i686 i386 GNU/Linux

[ernestoalejo]

I've used hg bisect to find the change that starts reproducing the error. This one fails for me:

https://code.google.com/p/go/source/detail?r=ddb9e6365e57

I'll investigate further

[mattn]

Could you please file this issue to https://code.google.com/p/go/issues/list ?

[ernestoalejo]

https://code.google.com/p/go/issues/detail?id=4935

[mattn]

Thank you.

[ernestoalejo]

It's now fixed:
https://code.google.com/p/go/issues/detail?id=4902
https://code.google.com/p/go/source/detail?r=3486a9feedda

[mattn]

Thank you.