On (18/09/08 13:44), Chase Douglas wrote: > On Sep 18, 2008, at 2:36 AM, Khem Raj wrote: >> On (16/09/08 08:16), Chase Douglas wrote: >>> On Sep 16, 2008, at 3:22 AM, Bernhard Reutner-Fischer wrote: >>> >>>> On Mon, Sep 15, 2008 at 05:30:46PM -0400, Chase Douglas wrote: >>>>> Hello all, >>>>> >>>>> I submitted this as a bug report on uclibc.org >>>>> (http://busybox.net/bugs/view.php?id=4994). I wanted to be sure >>>>> that >>>>> it >>>> >>>> Can you reproduce it with current trunk? >>> >>> I tried to build up a toolchain using the current trunk but I ran >>> into >>> issues with one of the pthread headers. I can't remember specifically >>> what files were at issue, but it had something to do with the uClibc >>> headers including pthread.h which clobbered functions in auth-pam.c >>> in >>> openssh. If it is REALLY necessary that I test it on the current >>> trunk I >>> may be able to fix it up enough to build what's necessary just for >>> testing, but I'll really need to be able to backport any changes >>> anyways. >> >> I tried this testcase on arm NPTL on uclibc-nptl branch and it seemed >> to exit properly. >> Following is spitted out. Is that what is expected ? >> >> [EMAIL PROTECTED]:~# ./a.out >> Hung Thread Test: 30 threads @ 50 forks/loops >> system() without pipes not using /tmp >> hiding output >> ......................................................................................................................................................(2)(6)(5)(0)(8)(1)(14)(4)(17)(7)(3)(27)(28)(18)(25)(19)(21)(9)(10)(29)(20)(24)(15)(11)(13)(16)(12)(23)(22)(26) >> => Done <= > > There's a couple issues here. First is that we are not using the nptl > branch and are not interested in it due to code maturity reasons.
Understood. However I am interested in nptl and thought that your testcase is interesting to be tested on nptl. this reveals an issue in linuxthreads.old as it works on nptl. I am hoping that nptl is going to be more prominent in future of uclibc for similar reasons. Thus, > we are using the linuxthreads.old library available in the standard > uClibc. > > Secondly, please try the test again as "./a.out fork pipes". We see the > application progress without issue using the default arguments, but not > when we start to use forks and pipes. That seems to work well too here it the tail of the messages. .... Thread 24 Loop 0 Error: Success (24)Thread 21 Loop 0 Error: Success (21)Thread 23 Loop 0 Error: Success (23)Thread 22 Loop 0 Error: Success (22)Thread 26 Loop 0 Error: Success (26)Thread 25 Loop 0 Error: Success (25)Thread 27 Loop 0 Error: Success (27)Thread 29 Loop 0 Error: Success Thread 28 Loop 0 Error: Success (28)(29) => Done <= Thx -Khem _______________________________________________ uClibc mailing list [email protected] http://busybox.net/cgi-bin/mailman/listinfo/uclibc
