Quoting Garrett Cooper ([email protected]): > On Wed, Apr 28, 2010 at 1:54 AM, Munipradeep Beerakam > <[email protected]> wrote: > > Hi, > > > > I noticed that mqns test cases are failing. I am using March 2010 LTP. > > After changing mq_open syscall in mqns_01.c file, test is passing. Same > > thing is happening w.r.t. other mqns test cases. Below is the change I > > made. > > > > Changed > > mqd = syscall(__NR_mq_open, SLASH_MQ1, O_RDWR|O_CREAT|O_EXCL,0777, > > NULL); > > > > to > > mqd = mq_open(SLASH_MQ1, O_RDWR|O_CREAT|O_EXCL, 0777, NULL); > > > > Failure output I got is below: > > # ./mqns_01 > > posixmq_namespace_01 0 TINFO : Testing posix mq namespaces through > > unshare(2). > > > > mq_open: Permission denied > > posixmq_namespace_01 1 TFAIL : mq_open failed > > > > After changing as above, I got the below output: > > # ./mqns_01 > > posixmq_namespace_01 0 TINFO : Testing posix mq namespaces through > > unshare(2). > > > > posixmq_namespace_01 0 TINFO : Checking namespaces isolation from > > parent to child > > > > posixmq_namespace_01 1 TPASS : child process didn't find mqueue > > > > So, my question is that what is the difference between the two, i.e., > > the previous one and the changed one. Both are calling the same mq_open
See http://www.mail-archive.com/[email protected]/msg09073.html for the ansswer. > > system call. But first one is failing whereas the second one is > > succeeding. > > > > So, do we need to modify all mqns test cases to the new form? > > Track down the execution path by first starting with glibc. It > might be a library vs kernel mismatch or a bug. Make sure that the > syscall number matches per your architecture, determine whether or not > your architecture has issues executing code in a biarch manner (64-bit > on 32-bit) if it applies. > HTH, -serge ------------------------------------------------------------------------------ _______________________________________________ Ltp-list mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ltp-list
