On 14 August 2012 02:01, Dietmar Stölting <dietmar.stoelt...@t-online.de> wrote: > with this new syscall.c content above things are going in the right > direction:-). > I make a test with strace from the program testthread of the Qemu testsuite. > When I understand the result right, > threading works now with this new compiled qemu-i386. > The child and the parents tidptr NOW have the same number in one thread, and > different but also same in other thread. > This means for the not working program testclone: The functioncall with its > sets of parameters is just wrong there. > When you do a function call with those Flags as in testthread, threads can be > builded with qemu-i386. > So, the error is in the wrong calling of the function clone(). This can be > corrected. Please tell me your thoughts,
Yes, as I said, we know that threading does not work for i386 targets (it is also broken in more subtle ways for other targets). This is not going to get fixed until it is investigated by somebody who has the time and expertise with both i386 architecture and QEMU internals to produce a coherent fix which addresses all the problems in this area. (See also my remarks in comment #47 of bug 739785.) I'm sorry if that sounds a bit negative, but there is a reason this bug has been unfixed for over a year -- it's not a trivial one to fix, and it's not easy to evaluate whether a small patch is a component of the complete correct solution without investing the time to think about the problem as a whole. -- PMM