Original Helgrind report:
==14160== Thread #57: lock order "0x47F6B90 before 0x7C25A28" violated
==14160== at 0x4C27730: pthread_mutex_lock (in /.../valgrind-3.6.0/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==14160== by 0x4EFF87E: do_thread_exit (threads.c:664)
==14160== by 0x4E87B89: c_body (continuations.c:512)
==14160== by 0x4F16C21: vm_regular_engine (vm-i-system.c:960)
==14160== by 0x4E90F92: scm_call_4 (eval.c:506)
==14160== by 0x4E88372: scm_i_with_continuation_barrier (continuations.c:450)
==14160== by 0x4E88424: scm_c_with_continuation_barrier (continuations.c:546)
==14160== by 0x51CA3AF: GC_call_with_gc_active (pthread_support.c:1128)
==14160== by 0x4EFF2E0: with_guile_and_parent (threads.c:206)
==14160== by 0x51C46B4: GC_call_with_stack_base (misc.c:1505)
==14160== by 0x4EFF447: scm_with_guile (threads.c:917)
==14160== by 0x51C46B4: GC_call_with_stack_base (misc.c:1505)
==14160== Required order was established by acquisition of lock at 0x47F6B90
==14160== at 0x4C27730: pthread_mutex_lock (in /.../valgrind-3.6.0/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==14160== by 0x4F00262: fat_mutex_lock (threads.c:1362)
==14160== by 0x4F004DA: scm_lock_mutex_timed (threads.c:1462)
==14160== by 0x4F16C09: vm_regular_engine (vm-i-system.c:898)
==14160== by 0x4E90F4D: scm_call_3 (eval.c:499)
==14160== by 0x4EFFA4C: really_launch (threads.c:975)
==14160== by 0x4E87B89: c_body (continuations.c:512)
==14160== by 0x4F16C21: vm_regular_engine (vm-i-system.c:960)
==14160== by 0x4E90F92: scm_call_4 (eval.c:506)
==14160== by 0x4E88372: scm_i_with_continuation_barrier (continuations.c:450)
==14160== by 0x4E88424: scm_c_with_continuation_barrier (continuations.c:546)
==14160== by 0x4EFF289: with_guile_and_parent (threads.c:874)
==14160== followed by a later acquisition of lock at 0x7C25A28
==14160== at 0x4C27730: pthread_mutex_lock (in /.../valgrind-3.6.0/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==14160== by 0x4F00352: fat_mutex_lock (threads.c:1374)
==14160== by 0x4F004DA: scm_lock_mutex_timed (threads.c:1462)
==14160== by 0x4F16C09: vm_regular_engine (vm-i-system.c:898)
==14160== by 0x4E90F4D: scm_call_3 (eval.c:499)
==14160== by 0x4EFFA4C: really_launch (threads.c:975)
==14160== by 0x4E87B89: c_body (continuations.c:512)
==14160== by 0x4F16C21: vm_regular_engine (vm-i-system.c:960)
==14160== by 0x4E90F92: scm_call_4 (eval.c:506)
==14160== by 0x4E88372: scm_i_with_continuation_barrier (continuations.c:450)
==14160== by 0x4E88424: scm_c_with_continuation_barrier (continuations.c:546)
==14160== by 0x4EFF289: with_guile_and_parent (threads.c:874)
* libguile/threads.c (fat_mutex_lock): In the `m->level == 0' case,
release M's lock before taking T's `admin_mutex'.
* configure.ac: Add a check for GC_get_suspend_signal().
* libguile/scmsigs.c (GC_get_suspend_signal): Define a fallback
implementation if one isn't available.
(signal_delivery_thread): Unmask the suspend signal so that GC can
stop the world. Fixes test-pthread-create on libgc 7.1 and earlier.
Thanks to Frank Terbeck <ft@bewatermyfriend.org> for the report.
* libguile/strings.c (scm_i_allocate_string_pointers): Encode strings
using the current locale. Previously, Latin-1 was used. Indirectly,
this affects the encoding of strings in `system*', `execl', `execlp',
`execle', `environ', and `dynamic-args-call'.
(scm_makfromstrs): In header comment, clarify that the C strings are
interpreted according to the current locale encoding.
* NEWS: Add NEWS entry.
Added pipe2 and open modules, to get O_CLOEXEC.
* libguile/Makefile.am (guile_filter_doc_snarfage$(EXEEXT)): Add gnulib
here, in the native build case, for rpl_fflush if needed.
foo
* libguile/threads.h:
* libguile/threads.c (scm_i_reset_fluid): New internal function, resets
the binding of a fluid for all threads. Needed for fluid GC.
* libguile/fluids.c (new_fluid): Call scm_i_reset_fluid here.
* test-suite/tests/weaks.test: Add tests.
* libguile/hashtab.c (scm_hash_fn_set_x): Fix updates to weak-value hash
tables to not deadlock inside the alloc lock.
* libguile/expand.h:
* module/language/tree-il.scm: Rename "sequence" to "seq", and instead
of taking a list of expressions, take a head and a tail.
* module/language/tree-il/analyze.scm:
* module/language/tree-il/compile-glil.scm:
* module/language/tree-il/fix-letrec.scm:
* module/language/tree-il/spec.scm:
* module/language/elisp/compile-tree-il.scm:
* module/ice-9/psyntax.scm:
* module/ice-9/psyntax-pp.scm:
* module/ice-9/eval.scm:
* libguile/memoize.h:
* libguile/memoize.c:
* libguile/expand.c:
* libguile/eval.c: Adapt to the new seq format.
* libguile/_scm.h (SCM_OBJCODE_ENDIANNESS_OFFSET):
(SCM_OBJCODE_WORD_SIZE_OFFSET): New defines.
* libguile/objcodes.c (scm_write_objcode): Use target-endianness and
target-word-size when writing the objcode cookie.
* libguile/hashtab.c (scm_internal_hash_fold): Don't try to unlink
deleted weak pairs. Our previous code was buggy (`prev' should have
only been updated in the case of a successful traversal), but more
than that, we're not in the alloc lock.
Thanks very much to Michael Wells for the report, and the debugging!
* libguile/pthread-threads.h: As an experiment, just define
scm_i_pthread_create et al to their pthread_* functions, instead of
the GC_ ones. It seems sufficient, given that all uses of Guile have
to go through scm_with_guile anyway... Passes tests on my machine.
Committing to give buildbots a go at it.
* libguile/scmsigs.c (read_signal_pipe_data, signal_delivery_thread):
Leave guile when reading from the signal pipe. Hopefully that lets GC
know not to wake up this thread.
* libguile/smob.h (scm_new_smob, scm_new_double_smob): New constructors,
which do what SCM_NEWSMOB / SCM_NEWSMOB3 had done, but with inline
functions instead of macros. They also bail to scm_i_new_smob /
scm_i_new_double_smob in either the mark or the free case, so that the
inline definition doesn't reference other internal details like libgc
stuff.
(SCM_SMOB_TYPE_MASK et al): Move definitions up so the new_smob see
them as already being declared.
(SCM_NEWSMOB, SCM_RETURN_NEWSMOB, SCM_NEWSMOB2, SCM_RETURN_NEWSMOB2):
(SCM_NEWSMOB3, SCM_RETURN_NEWSMOB3): Reimplement in terms of the new
inline functions.
(scm_i_finalize_smob): Remove declaration, as it's no longer needed or
used.
Remove now-unneeded bdw-gc include.
* libguile/smob.c (finalize_smob): Rename from scm_i_finalize_smob, and
make static.
(scm_i_new_smob, scm_i_new_double_smob): Slow-path allocators.
* libguile/inline.c: Include smob.h, so as to reify scm_new_smob and
scm_new_double_smob.
* libguile/inline.h:
* libguile/gc.h (scm_cell, scm_double_cell, scm_words): Move
declarations and definitions back here, from inline.h. It's more
natural.
* libguile/inline.c: Include gc.h as well.
* libguile/__scm.h (SCM_C_EXTERN_INLINE): Move this definition here,
from inline.h. We'd like to support inline function definitions in
more header files: not just inline.h.
(SCM_CAN_INLINE, SCM_INLINE, SCM_INLINE_IMPLEMENTATION): New
definitions.
* libguile/gc.h (SCM_GC_MALLOC, SCM_GC_MALLOC_POINTERLESS): Define these
wrappers, which redirect to the GC_MALLOC macros when building Guile,
and the scm_gc_malloc functions otherwise. A step towards getting
BDW-GC out of Guile's API.
* libguile/inline.h: Simplify, using SCM_INLINE,
SCM_INLINE_IMPLEMENTATION, and SCM_IMPLEMENT_INLINES. Also use the
new SCM_GC_MALLOC macros.
* libguile/inline.h:
* libguile/gc.c: Remove declaration and definition of unused
"scm_newcell_count" and "scm_newcell2_count". Since SCM_INTERNAL is
"extern", these symbols were not externally visible anyway, at least
under Linux or Windows.
* libguile/inline.h:
* libguile/deprecated.h:
* libguile/deprecated.c (scm_immutable_cell, scm_immutable_double_cell):
Deprecate these, as the GC_STUBBORN API doesn't do anything any more.
* libguile/strings.c (scm_i_c_make_symbol): Change the one use of
scm_immutable_double_cell to scm_double_cell.
* libguile/goops.c (scm_i_define_class_for_vtable): Lock around
additions to the vtable_class_map. Use is unlocked though; solving
GOOPS parallel-initialization would be a good thing.
* libguile/srcprop.h: Remove internal scm_source_whash declaration.
* libguile/srcprop.c (scm_i_set_source_properties_x)
(scm_i_has_source_properties): New helpers.
(scm_source_whash): Make static.
* libguile/read.c (scm_read_sexp): Remove register declarations here;
let's trust the compiler. Remove code to incrementally build up a
copy; instead let's let scm_i_set_source_properties_x handle copying
the expression if needed.
(scm_read_quote, scm_read_syntax): Use scm_i_set_source_properties_x.
(recsexpr): Remove this helper from 1996.
(scm_read_sharp_extension): Instead of trying to recursively label
sharp-read subforms with source properties, just label the outside
form and rely on the macro-expander to propagate it down.
* libguile/symbols.c (symbols_lock): Rename from intern_lock.
(lookup_interned_symbol, lookup_interned_latin1_symbol): Instead of
faith-based programming, just use the mutex. Though I haven't seen
this break, Ken is right!
* libguile/threads.c (lock_mutex_return_void, unlock_mutex_return_void):
New static functions that simply call scm_lock_mutex and
scm_unlock_mutex, respectively, but return void instead of SCM.
(scm_dynwind_lock_mutex): Pass unlock_mutex_return_void to
scm_dynwind_unwind_handler_with_scm, and lock_mutex_return_void to
scm_dynwind_rewind_handler_with_scm. Previously, we passed
scm_unlock_mutex and scm_lock_mutex (which return SCM), but the
scm_dynwind_* functions expect pointers to functions which return
void. When SCM is of type union, this changes the calling conventions
of the functions on some platforms (e.g. GCC 4.5.2 and 4.5.3 on x86).
* libguile/threads.c (lock_mutex_return_void, unlock_mutex_return_void):
New static functions that simply call scm_lock_mutex and
scm_unlock_mutex, respectively, but return void instead of SCM.
(scm_dynwind_lock_mutex): Pass unlock_mutex_return_void to
scm_dynwind_unwind_handler_with_scm, and lock_mutex_return_void to
scm_dynwind_rewind_handler_with_scm. Previously, we passed
scm_unlock_mutex and scm_lock_mutex (which return SCM), but the
scm_dynwind_* functions expect pointers to functions which return
void. When SCM is of type union, this changes the calling conventions
of the functions on some platforms (e.g. GCC 4.5.2 and 4.5.3 on x86).
* libguile/__scm.h: Move all the SCM_WTA and SCM_GASSERT macros out of
here. Also remove the scm_call_generic declarations.
* libguile/deprecated.h (SCM_WTA_DISPATCH_0, SCM_WTA_DISPATCH_1):
(SCM_WTA_DISPATCH_2, SCM_WTA_DISPATCH_N): Deprecate. See below for
their replacements.
(SCM_GASSERT0, SCM_GASSERT1, SCM_GASSERT2, SCM_GASSERTn): Deprecate
these too.
(SCM_WTA_DISPATCH_1_SUBR): Deprecate this strange thing.
(scm_call_generic_0, scm_call_generic_1, scm_call_generic_2):
(scm_call_generic_3, scm_apply_generic): Remove, indicating their
replacements.
* libguile/print.c (iprin1):
* libguile/eq.c (scm_equal_p): Use scm_call_2 instead of
scm_call_generic_2.
* libguile/goops.h:
* libguile/goops.c: Remove scm_{call,apply}_generic definitions.
(scm_wta_dispatch_0, scm_wta_dispatch_1, scm_wta_dispatch_2):
(scm_wta_dispatch_n): New procedures, replacing the SCM_WTA macros.
* libguile/numbers.c (scm_lcm):
* libguile/procs.c (scm_setter): Remove uses of SCM_GASSERT.
* libguile/numbers.c (scm_lcm):
* libguile/procs.c (scm_setter):
* libguile/vectors.c: Use the procedural scm_wta routines instead of the
SCM_WTA macros.
