| commit | eddf49be5aee76e1feae3e5aa3696f0c307dc068 | |
| author | Alastair Bridgewater <nyef@kana.lisphacker.com> | |
| Sun, 28 Oct 2012 14:13:21 +0000 (28 10:13 -0400) | ||
| committer | Alastair Bridgewater <nyef@kana.lisphacker.com> | |
| Sat, 10 May 2014 19:08:39 +0000 (10 15:08 -0400) | ||
| tree | 608e6e5da8072bcbb4e59e66be86a1f57da460e6 | treesnapshot (tar.gz zip) |
| parent | 593fd1dcb8b29bf04febcfbc1865d4eafabc9d99 | commitdiff |
ARM PSEUDO-ATOMIC fixes.
* At several points in PSEUDO-ATOMIC I neglected to unquote the
various inputs to the macro, leading to build failures when the
macro was used.
* The static symbols used for PSEUDO-ATOMIC were not being
initialized, remaining unbound by default, leading to the runtime
thinking that the system is in P-A until the end of the first P-A
block... And then executing a syscall (the number of which is
(ASH UNBOUND-MARKER-WIDETAG (- N-FIXNUM-TAG-BITS)), and that I've
never bothered to look up) which clobbers at least R0 and possibly
other registers. Fixed, by initializing both to NIL during
startup if *PSEUDO-ATOMIC-INTERRUPTED* is a static symbol.
* At several points in PSEUDO-ATOMIC I neglected to unquote the
various inputs to the macro, leading to build failures when the
macro was used.
* The static symbols used for PSEUDO-ATOMIC were not being
initialized, remaining unbound by default, leading to the runtime
thinking that the system is in P-A until the end of the first P-A
block... And then executing a syscall (the number of which is
(ASH UNBOUND-MARKER-WIDETAG (- N-FIXNUM-TAG-BITS)), and that I've
never bothered to look up) which clobbers at least R0 and possibly
other registers. Fixed, by initializing both to NIL during
startup if *PSEUDO-ATOMIC-INTERRUPTED* is a static symbol.
| src/compiler/arm/macros.lisp | diffblobblamehistory | |
| src/runtime/thread.c | diffblobblamehistory |