* Specific problems in IA64 make check
** test-unwind
Representation of the relevant dynamic context:
non-rewindable
catch frame make cont.
o----o-----a----------b-------------c
\
\ call cont.
o-----o-----------d
A continuation is captured at (c), with a non-rewindable frame in the
dynamic context at (b). If a rewind through that frame was attempted,
Guile would throw to the catch at (a). Then the context unwinds back
past (a), then winds forwards again, and the captured continuation is
called at (d).
We should end up at the catch at (a). On ia64, we get an "illegal
instruction".
The problem is that Guile does not restore the ia64 register backing
store (RBS) stack (which is saved off when the continuation is
captured) until all the unwinding and rewinding is done. Therefore,
when the rewind code (scm_i_dowinds) hits the non-rewindable frame at
(b), the RBS stack hasn't yet been restored. The throw finds the
jmp_buf (for the catch at (a)) correctly from the dynamic context, and
jumps back to (a), but the RBS stack is invalid, hence the illegal
instruction.
This could be fixed by restoring the RBS stack earlier, at the same
point (copy_stack) where the normal stack is restored. But that
causes a problem in the next test...
** continuations.test
The dynamic context diagram for this case is similar:
non-rewindable
catch frame make cont.
a----x-----o----------b-------------c
\
\ call cont.
o-------d
The only significant difference is that the catch point (a) is
upstream of where the dynamic context forks. This means that the RBS
stack at (d) already contains the correct RBS contents for throwing
back to (a), so it doesn't matter whether the RBS stack that was saved
off with the continuation gets restored.
This test passes with the Guile 1.8.4 code, but fails (with an
"illegal instruction") when the code is changed to restore the RBS
stack earlier as described above.
The problem now is that the RBS stack is being restored _too_ early;
specifically when there is still stuff to do that relies on the old
RBS contents. When a continuation is called, the sequence of relevant
events is:
(1) Grow the (normal) stack until it is bigger than the (normal)
stack saved off in the continuation. (scm_dynthrow, grow_stack)
(2) scm_i_dowinds calls itself recursively, such that
(2.1) for each rewind (from (x) to (c)) that will be needed,
another frame is added to the stack (both normal and RBS),
with local variables specifying the required rewind; the
rewinds don't actually happen yet, they will happen when
the stack unwinds again through these frames
(2.2) required unwinds - back from where the continuation was
called (d) to the fork point (x) - are done immediately.
(3) The normal (i.e. non-RBS) stack that was stored in the
continuation is restored (i.e. copied on top of the actual
stack).
Note that this doesn't overwrite the frames that were added in
(2.1), because the growth in (1) ensures that the added frames
are beyond the end of the restored stack.
(4) ? Restore the RBS stack here too ?
(5) Return (from copy_stack) through the (2.1) frames, which means
that the rewinds now happen.
(6) setcontext (or longjmp) to the context (c) where the
continuation was captured.
The trouble is that step (1) does not create space in the RBS stack in
the same kind of way that it does for the normal stack. Therefore, if
the saved (in the continuation) RBS stack is big enough, it can
overwrite the RBS of the (2.1) frames that still need to complete.
This causes an illegal instruction when we return through those frames
and try to perform the rewinds.
* Fix
The key to the fix is that the saved RBS stack only needs to be
restored at some point before the next setcontext call, and that doing
it as close to the setcontext call as possible will avoid bad
interactions with the pre-setcontext stack. Therefore we do the
restoration at the last possible point, immediately before the next
setcontext call.
The situation is complicated by there being two ways that the next
setcontext call can happen.
- If the unwinding and rewinding is all successful, the next
setcontext will be the one from step (6) above. This is the
"normal" continuation invocation case.
- If one of the rewinds throws an error, the next setcontext will
come from the throw implementation code. (And the one in step (6)
will never happen.) This is the rewind error case.
In the rewind error case, the code calling setcontext knows nothing
about the continuation. So to cover both cases, we:
- copy (in step (4) above) the address and length of the
continuation's saved RBS stack to the current thread state
(SCM_I_CURRENT_THREAD)
- modify all setcontext callers so that they check the current
thread state for a saved RBS stack, and restore it if so before
calling setcontext.
* Notes
** I think rewinders cannot rely on using any stack data
Unless it can be guaranteed that the data won't go into a register.
I'm not 100% sure about this, but I think it follows from the fact
that the RBS stack is not restored until after the rewinds have
happened.
Note that this isn't a regression caused by the current fix. In Guile
1.8.4, the RBS stack was restored _after_ the rewinds, and this is
still the case now.
** Most setcontext calls for `throw' don't need to change the RBS stack
In the absence of continuation invocation, the setcontext call in the
throw implementation code always sets context to a place higher up the
same stack (both normal and RBS), hence no stack restoration is
needed.
* Other changes
** Using setcontext for all non-local jumps (for __ia64__)
Along the way, I read a claim somewhere that setcontext was more
reliable than longjmp, in cases where the stack has been manipulated.
I don't now have any reason to believe this, but it seems reasonable
anyway to leave the __ia64__ code using getcontext/setcontext, instead
of setjmp/longjmp.
(I think the only possible argument against this would be performance -
if getcontext was significantly slower than setjmp. It that proves to
be the case, we should revisit this.)
** Capping RBS base for non-main threads
Somewhere else along the way, I hit a problem in GC, involving the RBS
stack of a non-main thread. The problem was, in
SCM_MARK_BACKING_STORE, that scm_ia64_register_backing_store_base was
returning a value that was massively greater than the value of
scm_ia64_ar_bsp, leading to a seg fault. This is because the
implementation of scm_ia64_register_backing_store_base is only valid
for the main thread. I couldn't find a neat way of getting the true
RBS base of a non-main thread, but one idea is simply to call
scm_ia64_ar_bsp when guilifying a thread, and use the value returned
as an upper bound for that thread's RBS base. (Note that the RBS
stack grows upwards.)
(Were it not for scm_init_guile, we could be much more definitive
about this. We could take the value of scm_ia64_ar_bsp as a
definitive base address for the part of the RBS stack that Guile cares
about. We could also then discard
scm_ia64_register_backing_store_base.)
* threads.c (SCM_MARK_BACKING_STORE): Use scm_ia64_ar_bsp() and
scm_ia64_register_backing_store_base() instead of Linux-specific
implementations.
* gc.h (scm_ia64_register_backing_store_base, scm_ia64_ar_bsp):
New declarations.
* gc.c (__libc_ia64_register_backing_store_base): Declaration
removed.
(scm_ia64_register_backing_store_base, scm_ia64_ar_bsp): New, with
implementations for Linux and HP-UX.
* coop-pthreads.c (SCM_MARK_BACKING_STORE): Use scm_ia64_ar_bsp()
and scm_ia64_register_backing_store_base() instead of
Linux-specific implementations.
* continuations.h (__libc_ia64_register_backing_store_base):
Declaration removed.
(scm_t_contregs): New "fresh" field.
* continuations.c (ia64_getcontext): Removed.
(scm_make_continuation): Use continuation fresh field instead of
interpreting getcontext return values (which isn't portable). Use
scm_ia64_ar_bsp() and scm_ia64_register_backing_store_base()
instead of Linux-specific implementations.
(copy_stack_and_call): Use scm_ia64_register_backing_store_base()
instead of Linux-specific implementation.
* _scm.h (__ia64__): Also detect __ia64.
scm_catch_with_pre_unwind_handler, scm_with_throw_handler): New.
* throw.c (SCM_JBPREUNWIND, SCM_SETJBPREUNWIND): New.
(struct pre_unwind_data): New, replaces struct lazy_catch.
(scm_c_catch): New, replaces scm_internal_catch as the primary
catch API for C code; adds pre-unwind handler support.
(scm_internal_catch): Now just a wrapper for scm_c_catch, for back
compatibility.
(tc16_pre_unwind_data, pre_unwind_data_print,
make_pre_unwind_data, SCM_PRE_UNWIND_DATA_P): Renamed from
"lazy_catch" equivalents.
(scm_c_with_throw_handler): New, replaces scm_internal_lazy_catch
as the primary C API for a "lazy" catch.
(scm_internal_lazy_catch): Now just a wrapper for
scm_c_with_throw_handler, for back compatibility.
(scm_catch_with_pre_unwind_handler): Renamed from scm_catch; adds
pre-unwind handler support.
(scm_catch): Now just a wrapper for
scm_catch_with_pre_unwind_handler, for back compatibility.
(scm_with_throw_handler): New.
(scm_lazy_catch): Update comment to say that the handler can
return, and what happens if it does.
(toggle_pre_unwind_running): New.
(scm_ithrow): When identifying the throw target, take running
flags into account. In general, change naming of things from
"lazy_catch" to "pre_unwind". When throwing to a throw handler,
don't unwind the dynamic context first. Add dynwind framing to
manage the running flag of a throw handler. If a lazy catch or
throw handler returns, rethrow the same exception again. Add
pre-unwind support to the normal catch case (SCM_JMPBUFP).
* root.c (scm_internal_cwdr): Add NULL args to
scm_i_with_continuation_barrier call.
* dynwind.c: Change comment mentioning lazy-catch to mention
pre-unwind data and throw handler also.
* continuations.h (scm_i_with_continuation_barrier): Add
pre-unwind handler args.
* continuations.c (scm_i_with_continuation_barrier): Add
pre-unwind handler args, and pass on to scm_c_catch (changed from
scm_internal_catch).
(c_handler): Remove scm_handle_by_message_noexit call.
(scm_c_with_continuation_barrier): Call
scm_i_with_continuation_barrier with scm_handle_by_message_noexit
as the pre-unwind handler.
(scm_handler): Remove scm_handle_by_message_noexit call.
(s_scm_with_continuation_barrier): Call
scm_i_with_continuation_barrier with scm_handle_by_message_noexit
as the pre-unwind handler.
print.c, ports.c, mallocs.c, hooks.c, hashtab.c, fports.c,
guardians.c, filesys.c, coop-pthreads.c, continuations.c: Use
scm_uintprint to print unsigned integers, raw heap words, and
adresses, using a cast to scm_t_bits to turn pointers into
integers.
part of copy_stack_and_call.
(copy_stack_and_call): Copy the stack after unwinding and before
rewinding.
(scm_dynthrow): Do not call scm_dowinds, this is now done by
copy_stack_and_call.
* continuations.c: Redeclaration of getcontext() via the
__asm__ ("getcontext") directive.
* continuations.h: Include <ucontext.h> instead of
<sys/ucontext.h>.
non-zero is returned from a port or smob free function.
(scm_malloc, scm_realloc, scm_strndup, scm_strdup,
scm_gc_register_collectable_memory,
scm_gc_unregister_collectable_memory, scm_gc_malloc,
scm_gc_realloc, scm_gc_free, scm_gc_strndup, scm_gc_strdup): New.
* backtrace.c, continuations.c, convert.i.c, coop-threads.c,
debug-malloc.c, dynl.c, environments.c, environments.h,
extensions.c, filesys.c, fports.c, gc.c, gc.h, gh_data.c, goops.c,
guardians.c, hooks.c, init.c, keywords.c, load.c, numbers.c,
ports.c, posix.c, procs.c, rdelim.c, regex-posix.c, root.c,
smob.c, stime.c, strings.c, struct.c, struct.h, symbols.c, unif.c,
vectors.c, weaks.c: Use scm_gc_malloc/scm_malloc and
scm_gc_free/free instead of scm_must_malloc and scm_must_free, as
appropriate. Return zero from smob and port free functions.
* debug-malloc.c (scm_malloc_reregister): Handle "old == NULL".
* fports.c (scm_setvbuf): Reset read buffer to saved values when
it is pointing to the putback buffer.
(SCM_NUM2{SIZE,PTRDIFF,SHORT,USHORT,BITS,UBITS,INT,UINT}[_DEF]):
new macros.
* unif.h: type renaming:
scm_array -> scm_array_t
scm_array_dim -> scm_array_dim_t
the old names are deprecated, all in-Guile uses changed.
* tags.h (scm_ubits_t): new typedef, representing unsigned
scm_bits_t.
* stacks.h: type renaming:
scm_info_frame -> scm_info_frame_t
scm_stack -> scm_stack_t
the old names are deprecated, all in-Guile uses changed.
* srcprop.h: type renaming:
scm_srcprops -> scm_srcprops_t
scm_srcprops_chunk -> scm_srcprops_chunk_t
the old names are deprecated, all in-Guile uses changed.
* gsubr.c, procs.c, print.c, ports.c, read.c, rdelim.c, ramap.c,
rw.c, smob.c, sort.c, srcprop.c, stacks.c, strings.c, strop.c,
strorder.c, strports.c, struct.c, symbols.c, unif.c, values.c,
vectors.c, vports.c, weaks.c:
various int/size_t -> size_t/scm_bits_t changes.
* random.h: type renaming:
scm_rstate -> scm_rstate_t
scm_rng -> scm_rng_t
scm_i_rstate -> scm_i_rstate_t
the old names are deprecated, all in-Guile uses changed.
* procs.h: type renaming:
scm_subr_entry -> scm_subr_entry_t
the old name is deprecated, all in-Guile uses changed.
* options.h (scm_option_t.val): unsigned long -> scm_bits_t.
type renaming:
scm_option -> scm_option_t
the old name is deprecated, all in-Guile uses changed.
* objects.c: various long -> scm_bits_t changes.
(scm_i_make_class_object): flags: unsigned long -> scm_ubits_t
* numbers.h (SCM_FIXNUM_BIT): deprecated, renamed to
SCM_I_FIXNUM_BIT.
* num2integral.i.c: new file, multiply included by numbers.c, used
to "templatize" the various integral <-> num conversion routines.
* numbers.c (scm_mkbig, scm_big2num, scm_adjbig, scm_normbig,
scm_copybig, scm_2ulong2big, scm_dbl2big, scm_big2dbl):
deprecated.
(scm_i_mkbig, scm_i_big2inum, scm_i_adjbig, scm_i_normbig,
scm_i_copybig, scm_i_short2big, scm_i_ushort2big, scm_i_int2big,
scm_i_uint2big, scm_i_long2big, scm_i_ulong2big, scm_i_bits2big,
scm_i_ubits2big, scm_i_size2big, scm_i_ptrdiff2big,
scm_i_long_long2big, scm_i_ulong_long2big, scm_i_dbl2big,
scm_i_big2dbl, scm_short2num, scm_ushort2num, scm_int2num,
scm_uint2num, scm_bits2num, scm_ubits2num, scm_size2num,
scm_ptrdiff2num, scm_num2short, scm_num2ushort, scm_num2int,
scm_num2uint, scm_num2bits, scm_num2ubits, scm_num2ptrdiff,
scm_num2size): new functions.
* modules.c (scm_module_reverse_lookup): i, n: int -> scm_bits_t.x
* load.c: change int -> size_t in various places (where the
variable is used to store a string length).
(search-path): call scm_done_free, not scm_done_malloc.
* list.c (scm_ilength): return a scm_bits_t, not long.
some other {int,long} -> scm_bits_t changes.
* hashtab.c: various [u]int -> scm_bits_t changes.
scm_ihashx_closure -> scm_ihashx_closure_t (and made a typedef).
(scm_ihashx): n: uint -> scm_bits_t
use scm_bits2num instead of scm_ulong2num.
* gsubr.c: various int -> scm_bits_t changes.
* gh_data.c (gh_scm2double): no loss of precision any more.
* gh.h (gh_str2scm): len: int -> size_t
(gh_{get,set}_substr): start: int -> scm_bits_t,
len: int -> size_t
(gh_<num>2scm): n: int -> scm_bits_t
(gh_*vector_length): return scm_[u]size_t, not unsigned long.
(gh_length): return scm_bits_t, not unsigned long.
* fports.h: type renaming:
scm_fport -> scm_fport_t
the old name is deprecated, all in-Guile uses changed.
* fports.c (fport_fill_input): count: int -> scm_bits_t
(fport_flush): init_size, remaining, count: int -> scm_bits_t
* debug.h (scm_lookup_cstr, scm_lookup_soft, scm_evstr): removed
those prototypes, as the functions they prototype don't exist.
* fports.c (default_buffer_size): int -> size_t
(scm_fport_buffer_add): read_size, write_size: int -> scm_bits_t
default_size: int -> size_t
(scm_setvbuf): csize: int -> scm_bits_t
* fluids.c (n_fluids): int -> scm_bits_t
(grow_fluids): old_length, i: int -> scm_bits_t
(next_fluid_num, scm_fluid_ref, scm_fluid_set_x): n: int ->
scm_bits_t
(scm_c_with_fluids): flen, vlen: int -> scm_bits_t
* filesys.c (s_scm_open_fdes): changed calls to SCM_NUM2LONG to
the new and shiny SCM_NUM2INT.
* extensions.c: extension -> extension_t (and made a typedef).
* eval.h (SCM_IFRAME): cast to scm_bits_t, not int. just so
there are no nasty surprises if/when the various deeply magic tag
bits move somewhere else.
* eval.c: changed the locals used to store results of SCM_IFRAME,
scm_ilength and such to be of type scm_bits_t (and not int/long).
(iqq): depth, edepth: int -> scm_bits_t
(scm_eval_stack): int -> scm_bits_t
(SCM_CEVAL): various vars are not scm_bits_t instead of int.
(check_map_args, scm_map, scm_for_each): len: long -> scm_bits_t
i: int -> scm_bits_t
* environments.c: changed the many calls to scm_ulong2num to
scm_ubits2num.
(import_environment_fold): proc_as_ul: ulong -> scm_ubits_t
* dynwind.c (scm_dowinds): delta: long -> scm_bits_t
* debug.h: type renaming:
scm_debug_info -> scm_debug_info_t
scm_debug_frame -> scm_debug_frame_t
the old names are deprecated, all in-Guile uses changed.
(scm_debug_eframe_size): int -> scm_bits_t
* debug.c (scm_init_debug): use scm_c_define instead of the
deprecated scm_define.
* continuations.h: type renaming:
scm_contregs -> scm_contregs_t
the old name is deprecated, all in-Guile uses changed.
(scm_contregs_t.num_stack_items): size_t -> scm_bits_t
(scm_contregs_t.num_stack_items): ulong -> scm_ubits_t
* continuations.c (scm_make_continuation): change the type of
stack_size form long to scm_bits_t.
* ports.h: type renaming:
scm_port_rw_active -> scm_port_rw_active_t (and made a typedef)
scm_port -> scm_port_t
scm_ptob_descriptor -> scm_ptob_descriptor_t
the old names are deprecated, all in-Guile uses changed.
(scm_port_t.entry): int -> scm_bits_t.
(scm_port_t.line_number): int -> long.
(scm_port_t.putback_buf_size): int -> size_t.
* __scm.h (long_long, ulong_long): deprecated (they pollute the
global namespace and have little value besides that).
(SCM_BITS_LENGTH): new, is the bit size of scm_bits_t (i.e. of an
SCM handle).
(ifdef spaghetti): include sys/types.h and sys/stdtypes.h, if they
exist (for size_t & ptrdiff_t)
(scm_sizet): deprecated.
* Makefile.am (noinst_HEADERS): add num2integral.i.c
* values.c, values.h: new files. use a struct to contain multiple
values, similar to the previous Scheme-level implementation.
* Makefile.am: add values.c, values.h, values.x.
* continuations.c (continuation_apply): support R5RS multiple value
continuations.
* init.c: call scm_init_values.
* struct.h: define SCM_SET_STRUCT_PRINTER.
custom tc7 type. This will make it easier to support R5RS
multiple value continuations, without the use of a Scheme-level
wrapper.
* continuations.c (scm_tc16_continuation, continuation_mark,
continuation_free, continuation_print, continuation_apply):
new SMOB support.
(scm_make_continuation): new procedure, replaces scm_make_cont
with a different interface.
(copy_stack_and_call, scm_dynthrow, scm_init_continuations): rewritten.
(CHEAP_CONTINUATIONS): removed non-working code completely.
(scm_call_continuation): removed.
* continuations.h (struct scm_contregs): add num_stack_items and
stack fields. previously stack was stored following this struct:
use a tail array instead.
(SCM_CONTINUATIONP): new macro.
(SCM_CONTINUATION_LENGTH, SCM_SET_CONTINUATION_LENGTH):
rewritten.
(SCM_SET_CONTREGS): removed.
* tags.h: removed scm_tc7_contin (was tag 61).
* debug.c, gc.c, hash.c, print.c, procprop.c, procs.c:
removed scm_tc7_contin support.
* eval.c: use scm_make_continuation instead of scm_make_cont.
don't set jump buffers here. remove scm_tc7_contin support.
* init.c, root.c: create SMOB continuation for rootcont instead
of scm_tc7_contin. call scm_init_continuations before
scm_init_root.
* root.c: remove support for static jmpbuf. It's not used by
default and I broke it. create SMOB continuation for rootcont.
* stacks.c: use SCM_CONTINUATIONP.
and we don't want that.
(INCLUDES): Removed all -I options except for the root source
directory and the root build directory.
* *.*: Change includes so that they always use the "prefixes"
libguile/, qt/, guile-readline/, or libltdl/.
(Thanks to Tim Mooney.)
