Since tile has not yet been supported in an official release, this bug is just a placeholder for bug fixes that have been made in master during the runup to the 2.16 release. The seven correctness bugs that have been fixed should be backported to 2.15. (Cleanup, performance bugs, and purely test-support changes don't, I think, rise to the level of being necessary to backport.) commit 3c19868059809171740f3f71ec0e950582b3b23b Author: Chris Metcalf <cmetcalf@tilera.com> Date: Fri May 11 15:57:05 2012 -0400 linux-generic: add wordsize-32 version of prlimit64/fanotify syscalls commit 575298fcd298bb3ff7d7ba0d0bceac9429ae2b5d Author: Chris Metcalf <cmetcalf@tilera.com> Date: Fri May 11 17:59:23 2012 -0400 tile: allow memcpy(p, p, n) without corrupting memory at "p" Although this is not required by the definition of memcpy(), in practice this sort of thing does happen, and it's easy to make the code robust by doing nothing in this case. (Since structure copy causes the compiler to emit a memcpy, in the case where the target structure is the same as the destination, we were seeing corruption.) commit 024bb7b4a22e62faf932a7b9f071322716727d2d Author: Chris Metcalf <cmetcalf@tilera.com> Date: Fri May 11 18:08:03 2012 -0400 tile: fix bug in CFI definitions for clone.S We were missing a critical cfi_def_cfa_offset call in .Lthread_start which caused backtracing to get badly confused for threads. commit 8927be2b217ad44682efd987e32320ee80c4dbe1 Author: Chris Metcalf <cmetcalf@tilera.com> Date: Fri May 11 18:10:00 2012 -0400 tile: add include <features.h> to <sys/dataplane.h> We were using __BEGIN_DECLS without <features.h> and it just happened to mostly work, but of course we should include it here. commit bcfe09d24161bdf3b734da2f2fc4692bd26b7f52 Author: Chris Metcalf <cmetcalf@tilera.com> Date: Mon May 14 15:03:46 2012 -0400 tile: align stack for tilegx32 Previously we weren't re-aligning the stack pointer during the call to _dl_init(), so for tilegx32 and an odd value in _dl_skip_args and kernel unaligned access fixups disabled, we would die with SIGBUS. We now handle this case properly by aligning before calling _dl_init(). commit 0adc5f3892fee6fc6221691e3f269fcc3b386965 Author: Chris Metcalf <cmetcalf@tilera.com> Date: Wed May 16 13:52:36 2012 -0400 tilegx32: fix various bugs in setcontext/getcontext/swapcontext commit 30aa747c6895cfb6ae7c9655375c1d0c87acdc41 Author: Chris Metcalf <cmetcalf@tilera.com> Date: Thu May 17 15:18:45 2012 -0400 tile: add proper versioning for fegetenv()
Backported and tested.
Chris, Could you attach the patch to this issue? I can review the patch and OK you to check it into the stable 2.15 branch?
Created attachment 6464 [details] rollup of seven cherry-picked patches These patches are all on master already.
(In reply to comment #3) > Created attachment 6464 [details] > rollup of seven cherry-picked patches > > These patches are all on master already. Chris, Please split this into at *least* 2 commits, one for tile changes, one for linux-generic changes. OK to commit to 2.15 stable branch with those changes (and the expectation of no new regressions in the testsuite).
(In reply to comment #4) > Chris, > > Please split this into at *least* 2 commits, one for tile changes, one for > linux-generic changes. > > OK to commit to 2.15 stable branch with those changes (and the expectation of > no new regressions in the testsuite). I think we're actually fine. The changes in my repository for 2.15 are seven commits, cherry picked in the usual way. I haven't merged them into a single commit or anything like that, and I was planning to push them as 7 commits, including the "(cherry picked from commit blah blah)" lines in the commit messages. I haven't tested any platforms other than tilepro, tilegx64, and tilegx32, but these changes aren't seen by any other platforms (yet - perhaps more to come with linux-generic, but not yet that I know of). Let me know if this sounds like what you were expecting :-)
(In reply to comment #5) > (In reply to comment #4) > > Chris, > > > > Please split this into at *least* 2 commits, one for tile changes, one for > > linux-generic changes. > > > > OK to commit to 2.15 stable branch with those changes (and the expectation of > > no new regressions in the testsuite). > > I think we're actually fine. The changes in my repository for 2.15 are seven > commits, cherry picked in the usual way. I haven't merged them into a single > commit or anything like that, and I was planning to push them as 7 commits, > including the "(cherry picked from commit blah blah)" lines in the commit > messages. > > I haven't tested any platforms other than tilepro, tilegx64, and tilegx32, but > these changes aren't seen by any other platforms (yet - perhaps more to come > with linux-generic, but not yet that I know of). > > Let me know if this sounds like what you were expecting :-) That rationale is exactly what I was expecting, and sound good. Please commit to 2.15 branch.
As this was approved, you should commit it to the branch and close the bug as fixed....
Pushed.