RFR(S): 8208623: [TESTBUG] runtime/LoadClass/LongBCP.java fails in AUFS file system

[David Holmes]

On 2/08/2018 12:46 AM, Pengfei Li wrote:
> Hi David,
> 
> 
> I've tried to find such kind of Java APIs but didn't get an appropriate 
> one. Maybe the os file system type is designed to be transparent to Java 
> applications. So the only way I could think of is to execute a shell 
> command.

What does FileStore.type() show for AUFS? This little test seems to 
report other filesystem types okay:

import java.nio.file.*;

public class FileSystemCheck {
   public static void main(String[] args) throws Throwable {
     FileStore fs = Files.getFileStore(new java.io.File(args[0]).toPath());
     System.out.println(fs.type());
   }
}

 > java9 FileSystemCheck /java
autofs
 > java9 FileSystemCheck /
rootfs
 > java9 FileSystemCheck /proc
proc
 > java9 FileSystemCheck .
ext4

David
-----

> 
> --
> 
> Thanks,
> 
> Pengfei
> 
> ------------------------------------------------------------------------
> *From:* David Holmes <david.holmes at oracle.com>
> *Sent:* Wednesday, August 1, 2018 8:24:31 PM
> *To:* Pengfei Li; hotspot-runtime-dev at openjdk.java.net
> *Cc:* nd
> *Subject:* Re: RFR(S): 8208623: [TESTBUG] runtime/LoadClass/LongBCP.java 
> fails in AUFS file system
> Hi Pengfei,
> 
> On 1/08/2018 6:14 PM, Pengfei Li wrote:
>> Hi,
>> 
>> Could you please help review the small fix to jtreg test case 
>> “runtime/LoadClass/LongBCP.java” ?
>> 
>> This case fails in Linux AUFS file system since the maximum file name 
>> length is shorter than 255.
>> 
>> JBS: https://bugs.openjdk.java.net/browse/JDK-8208623
>> 
>> webrev: http://cr.openjdk.java.net/~zyao/8208623/webrev.00/index.html 
> <http://cr.openjdk.java.net/%7Ezyao/8208623/webrev.00/index.html>
>> <http://cr.openjdk.java.net/%7Ezyao/8208623/webrev.00/index.html>
> 
> I'm wondering whether the filesystem type can be found through one of
> the java.nio.file APIs i.e FileSystem/FileStore/FileSystemProvider so
> that you do not have to exec the "stat -f -c %T ." command ??
> 
> It would also be beneficial for checking the operation of the test to
> print the "max length" value that it ends up testing.
> 
> Thanks,
> David
> 
>> "Since aufs has several filename prefixes reserved, the maximum filename 
>> length is shorter than ordinary 255. Actually 242 (defined as 
>> ${AUFS_MAX_NAMELEN})." [1]
>> 
>> --
>> 
>> Thanks,
>> 
>> Pengfei
>> 
>> [1] http://aufs.sourceforge.net/aufs3/man.html
>>