I also have some problems in NTPL,. Some testcases work well in linuxthreads, but not in NPTL.
xinglong > -------Original Message------- > From: Qais Yousef <[email protected]> > To: Bernhard Reutner-Fischer <[email protected]> > Cc: uClibc <[email protected]> > Subject: RE: open() return value > Sent: Jul 17 '12 22:34 > > > -----Original Message----- > > From: [email protected] [mailto:[email protected]] On > Behalf > > Of Qais Yousef > > Sent: 17 July 2012 13:37 > > To: Bernhard Reutner-Fischer > > Cc: uClibc > > Subject: RE: open() return value > > > > > -----Original Message----- > > > From: Bernhard Reutner-Fischer [mailto:[email protected]] > > > Sent: 13 July 2012 21:45 > > > To: Qais Yousef > > > Cc: [email protected]; uClibc > > > Subject: RE: open() return value > > > > > > Can you reproduce which sequence of calls trigger this behaviour? Can > > > you reproduce it with current master? What arch do you use and which > > > version of uClibc? > > > Thanks, > > I could trigger the same problem on a simple test app by doing a couple of > open() > > to a non-existent file, create a thread then do the open() again. Too bad > I can't > > use master, I'm on 0.9.32.1 using our own arch. > > > > What's the difference between __uclibc_open() and normal open()? For me > > open.c isn't compiled at all and open64 should be used instead, AFAICT at > least. In > > my shared object both __uclibc_open() and open64() symbols are defined, and > > for the failed case __uclibc_open() gets called. I would have expected to > have 1 > > open function only? > > > > Sorry for the delayed response as I am working on this on and off. > > > > Cheers, > > Qais > > A small update. Using linuxthreads instead of NPTL everything works > correctly.. > > Qais > > _______________________________________________ > > uClibc mailing list > > [email protected] > > http://lists.busybox.net/mailman/listinfo/uclibc > > _______________________________________________ > uClibc mailing list > [email protected] > http://lists.busybox.net/mailman/listinfo/uclibc > _______________________________________________ uClibc mailing list [email protected] http://lists.busybox.net/mailman/listinfo/uclibc
