[PATCH] libdispatch-r197_v1
I have attached a patch against r197 that is a combination of several previous uncommitted patches, plus a few other things. With this patch, I can perform a simple "./configure && make" on Linux, and everything works. Regards, - Mark
On 06/09/2011 05:27 AM, Mark Heily wrote:
Description: multiple fixes for libdispatch r197 * Workaround a link-time error by avoiding the use of __private_extern__ on GNU/Linux.
Have you reported this? Does it happen with both the LLVM internal assembler/linker (I'm a little rusty, I don't remember if the LLVM linker exists for ELF or is only for Mac OS) and GNU binutils? Can you send (here or privately) the output of "make >error.log 2>&1" with the failure?
* Workaround a pthread_exit() bug on Linux that causes the process to become a zombie.
Can you detail this more? I remember seeing something on the list, but I am not sure. Paolo
On 06/09/2011 01:56 AM, Paolo Bonzini wrote:
On 06/09/2011 05:27 AM, Mark Heily wrote:
Description: multiple fixes for libdispatch r197 * Workaround a link-time error by avoiding the use of __private_extern__ on GNU/Linux.
Have you reported this? Does it happen with both the LLVM internal assembler/linker (I'm a little rusty, I don't remember if the LLVM linker exists for ELF or is only for Mac OS) and GNU binutils? Can you send (here or privately) the output of "make >error.log 2>&1" with the failure?
I haven't reported the problem. It doesn't happen with the LLVM linker, however llvm-ld cannot generate shared libraries on Linux; it can only produce static libraries. I've attached the error log as you requested.
* Workaround a pthread_exit() bug on Linux that causes the process to become a zombie.
Can you detail this more? I remember seeing something on the list, but I am not sure.
When the main thread calls pthread_exit() in Linux, the process becomes a zombie. My memory is fuzzy, but I seem to remember having problems sending SIGSTOP to the process, and was also unable to attach a debugger to the process. Here is an LKML thread that discusses the problem and references a glibc bug report: http://kerneltrap.org/mailarchive/linux-kernel/2009/2/1/4404553 I checked kernel/exit.c in Linux 2.3.35, and it appears the kernel patch to fix the issue has not been applied. Regards, - Mark
participants (2)
-
Mark Heily
-
Paolo Bonzini