As the Autoconf documentation says, "These days portable programs
[...] should not rely on `HAVE_RESTARTABLE_SYSCALLS', since nowadays
whether a system call is restartable is a dynamic issue, not a
configuration-time issue."
In other words, if we ever rely on HAVE_RESTARTABLE_SYSCALLS, we are
at the mercy of any code that Guile happens to be linked with, because
that code could install a signal handler without the SA_RESTART flag,
and then a Guile system call could unexpectedly return EINTR.
The readline part of this goes back to this problem report:
excellent example of the above paragraph. It was noted during the
discussion that undefining HAVE_RESTARTABLE_SYSCALLS would fix the
problem, but that solution wasn't adopted - I guess because Guile was
still using cooperative threads then (not pthreads) and so there was a
significant concern (whether founded or not) that not using
restartable syscalls (where available) could lead to a loss of
performance.
Now Guile's default mode of operation is with pthreads, where we
already don't assume that HAVE_RESTARTABLE_SYSCALLS is reliable, so
there is no possible further performance loss. And in any case we
really have no choice, if we want correct operation.
* configure.in (AC_SYS_RESTARTABLE_SYSCALLS): Removed.
* doc/ref/posix.texi (Signals): Remove statement that Guile always
sets SA_RESTART flag.
* guile-readline/configure.in (GUILE_SIGWINCH_SA_RESTART_CLEARED):
Remove this setting, together with its test code.
(HAVE_RL_PRE_INPUT_HOOK): Remove this setting and its code, as no
longer needed.
* guile-readline/readline.c (sigwinch_enable_restart): Removed.
(scm_init_readline): Remove setting of rl_pre_input_hook.
* libguile/_scm.h (SCM_SYSCALL): Remove the definition that relies on
HAVE_RESTARTABLE_SYSCALLS.
* libguile/scmsigs.c (scm_sigaction_for_thread): Don't always set the
SA_RESTART flag if available. Update docstring accordingly.
(scm_init_scmsigs): Remove code that sets SA_RESTART flag for all
signals.
Thanks to Greg Troxel for reporting, and Barry Fishman for the
explanation and fix.
* test-suite/tests/popen.test ("open-input-pipe"): Use shell function
`read' with an explicit argument, as apparently not all shells
support read with no argument.
* libguile/threads.c (scm_join_thread): Always recheck t->exited
before calling block_self again, in case thread t has now exited.
* test-suite/tests/threads.test (joining): New test.
On the one hand we want the child process in these tests to exit. On
the other, we don't want it to exit before the parent Guile code has
tested the relevant condition (EOF in the first test, broken pipe in
the second) - because these conditions would obviously be true if the
child had already exited, and that's not what we're trying to test
here. We're trying to test getting EOF and broken pipe while the
child process is still alive.
* test-suite/tests/popen.test (open-input-pipe:no duplicate): Add
another pipe from parent to child, so that the child can finish by
reading from this. Then the parent controls the child lifetime by
writing to this pipe.
* test-suite/tests/popen.test (open-output-pipe:no duplicate): Add
another pipe from child to parent, and have the child finish by
endlessly writing into this. Then the parent controls the child
lifetime by closing its end of the pipe, causing a broken pipe in
the child.
* test-suite/tests/reader.test ("read-options"): Add test
for long postfix keywords.
* libguile/read.c (scm_read_mixed_case_symbol): Fix
exception on symbols are greater than 128 chars. Also,
colons are not stripped from long postfix keywords.
* configure.in: Check for the declaration of flock(2).
* libguile/posix.c [!__MINGW32__][!HAVE_DECL_FLOCK]: Provide an flock(2)
declaration, needed on Tru64 5.1b.
* test-suite/standalone/test-round.c (HAVE_MACHINE_FPU_H): Include
<sys/types.h> when available. This fixes compilation on NetBSD.
Reported by Greg Toxel.
* configure.in (CPPFLAGS): Add `-mieee' or `-ieee' on Alpha and SH.
* libguile/numbers.c (guile_ieee_init): Make sure `-mieee' was passed
when using GCC on Alpha.
* NEWS: Update.
* test-suite/standalone/test-conversion.c (ieee_init): New function.
(guile_Inf, guile_NaN): New variables.
(test_from_double, test_to_double): Use them. Divisions by zero made
`cc' on Tru64 5.1b ("Compaq C V6.5-011") bail out and led to a
floating point exception when compiled with GCC on the same platform.
(main): Call `ieee_init ()'.
* configure.in: Check for the declaration of unsetenv(3), which Tru64
5.1b doesn't have.
* libguile/posix.c: Include <stdlib.h> since that's where POSIX says
unsetenv(3) should reside.
(scm_putenv): Don't attempt to use unsetenv(3) if its declaration
isn't available since that wouldn't work well on Tru64.
* test-suite/standalone/test-conversion.c: Remove extraneous semicolon
following `DEF[SU]TST' invocations since that made Compaq C
V6.5-011 (`cc' on Tru64 5.1b) bail out.
* configure.in (enable_deprecated): Set SCM_WARN_DEPRECATED_DEFAULT
even when --disable-deprecated is passed.
* libguile/deprecation.h: Declare deprecation-issuing methods even
if SCM_ENABLE_DEPRECATED is not set.
* libguile/deprecation.c: Ditto.
(scm_init_deprecation): Include full body even for unset
SCM_ENABLE_DEPRECATED.
Signed-off-by: Ludovic Courtès <ludo@gnu.org>
The crux of this problem was that the thread doing a throw, and so
checking scm_i_critical_section_level, was different from the thread
that was in a critical section.
* libguile/async.h (scm_i_critical_section_level): Removed, replaced
by per-thread critical_section_level.
(SCM_CRITICAL_SECTION_START, SCM_CRITICAL_SECTION_END): Use
per-thread critical_section_level.
* libguile/continuations.c (scm_dynthrow): Check per-thread
critical_section_level.
* libguile/threads.c (guilify_self_1): Init per-thread
critical_section_level.
(scm_i_critical_section_level): Removed.
* libguile/threads.h (scm_i_thread): New critical_section_level field.
* libguile/throw.c (scm_ithrow): Check per-thread critical_section_level.
* libguile/symbols.c (intern_symbol): New function, with code formerly
duplicated in `scm_i_c_mem2symbol ()' and `scm_i_mem2symbol ()'.
(scm_i_c_mem2symbol, scm_i_mem2symbol): Use it.
(scm_take_locale_symboln): Use `intern_symbol ()'. This fixes
bug #25865.
* test-suite/standalone/Makefile.am
(test_scm_take_locale_symbol_SOURCES,
test_scm_take_locale_symbol_CFLAGS,
test_scm_take_locale_symbol_LDADD): New variables.
(check_PROGRAMS, TESTS): Add `test-scm-take-locale-symbol'.
* benchmark-suite/benchmarks/subr.bm (hook1, hook3): New variables.
("subr invocation")("generic subr with rest arg", "generic subr with
rest arg and 3+ parameters"): New benchmarks.
("subr application")("generic subr with rest arg", "generic subr with
rest arg and 3+ parameters"): New benchmarks.
This fixes the following helgrind report.
Thread #1: lock order "0x4114748 before 0x4331084" violated
at 0x40234F7: pthread_mutex_lock (hg_intercepts.c:408)
by 0x407A55F: deval (eval.c:4077)
by 0x407A071: scm_dapply (eval.c:5012)
by 0x407888A: scm_apply (eval.c:4811)
by 0x407E20C: scm_call_0 (eval.c:4666)
by 0x406BDF7: scm_dynamic_wind (dynwind.c:111)
by 0x407620C: ceval (eval.c:4571)
by 0x407E8D9: scm_primitive_eval_x (eval.c:5921)
by 0x407E934: scm_eval_x (eval.c:5956)
by 0x40B9140: scm_shell (script.c:737)
by 0x4095AC5: invoke_main_func (init.c:367)
by 0x4065A81: c_body (continuations.c:349)
Required order was established by acquisition of lock at 0x4114748
at 0x40234F7: pthread_mutex_lock (hg_intercepts.c:408)
by 0x40B7BE1: scm_sigaction_for_thread (scmsigs.c:339)
by 0x40911DE: scm_gsubr_apply (gsubr.c:223)
by 0x4079E36: scm_dapply (eval.c:4930)
by 0x407AB68: deval (eval.c:4378)
by 0x407A071: scm_dapply (eval.c:5012)
by 0x407888A: scm_apply (eval.c:4811)
by 0x407E1D0: scm_call_1 (eval.c:4672)
by 0x407FEEE: scm_map (eval.c:5489)
by 0x407BDCC: deval (eval.c:4367)
by 0x407D197: deval (eval.c:3698)
by 0x407A071: scm_dapply (eval.c:5012)
followed by a later acquisition of lock at 0x4331084
at 0x40234F7: pthread_mutex_lock (hg_intercepts.c:408)
by 0x40DC397: scm_enter_guile (threads.c:377)
by 0x40DC8F4: scm_pthread_mutex_lock (threads.c:1485)
by 0x4083FAA: scm_gc_for_newcell (gc.c:484)
by 0x40A9781: scm_cons (inline.h:115)
by 0x4079867: scm_dapply (eval.c:4850)
by 0x407888A: scm_apply (eval.c:4811)
by 0x40796D6: scm_call_3 (eval.c:4684)
by 0x409A7C2: module_variable (modules.c:302)
by 0x409A7EE: module_variable (modules.c:312)
by 0x409A962: scm_sym2var (modules.c:466)
by 0x40738F4: scm_lookupcar1 (eval.c:2874)
* libguile/scmsigs.c (close_1): Renamed `handler_to_async'; also
handle #f case and wrapping the async in a cons, if necessary.
(install_handler): Pass in async instead of constructing it; combine
two branches into one.
(scm_sigaction_for_thread): Allocate async upfront instead of inside
the critical section, and pass it to install_handler calls. Leave
critical section before signaling out-of-range error.
This fixes lots of helgrind-reported problems, such as:
Thread #1: lock order "0x4325084 before 0x4108928" violated
at 0x40234F7: pthread_mutex_lock (hg_intercepts.c:408)
by 0x40CFD37: scm_enter_guile (threads.c:377)
by 0x40D0284: scm_pthread_mutex_lock (threads.c:1485)
by 0x405A736: scm_async_click (async.c:155)
by 0x406F9EE: deval (eval.c:4080)
by 0x40761D9: scm_primitive_eval_x (eval.c:5921)
by 0x40AD20E: install_handler (scmsigs.c:113)
by 0x40AD402: scm_sigaction_for_thread (scmsigs.c:394)
by 0x4087D1F: scm_gsubr_apply (gsubr.c:223)
by 0x406DF55: scm_dapply (eval.c:4930)
by 0x407147C: deval (eval.c:4378)
by 0x406E1BD: scm_dapply (eval.c:5012)
Required order was established by acquisition of lock at 0x4325084
at 0x40234F7: pthread_mutex_lock (hg_intercepts.c:408)
by 0x40CFD37: scm_enter_guile (threads.c:377)
by 0x408C58B: scm_i_init_guile (init.c:421)
by 0x40D1873: scm_i_init_thread_for_guile (threads.c:589)
by 0x40D18B4: scm_i_with_guile_and_parent (threads.c:731)
by 0x40D19BD: scm_with_guile (threads.c:720)
by 0x408C42E: scm_boot_guile (init.c:350)
by 0x8048710: main (guile.c:69)
followed by a later acquisition of lock at 0x4108928
at 0x40234F7: pthread_mutex_lock (hg_intercepts.c:408)
by 0x40AFD61: scm_make_smob_type (smob.c:294)
by 0x40AFF19: scm_smob_prehistory (smob.c:512)
by 0x408C595: scm_i_init_guile (init.c:423)
by 0x40D1873: scm_i_init_thread_for_guile (threads.c:589)
by 0x40D18B4: scm_i_with_guile_and_parent (threads.c:731)
by 0x40D19BD: scm_with_guile (threads.c:720)
by 0x408C42E: scm_boot_guile (init.c:350)
by 0x8048710: main (guile.c:69)
* libguile/async.c (scm_async_click): Don't leave Guile mode when
locking async_mutex. We don't need to, because none of the code
that has async_mutex locked can block, and doing so may lead to lock
ordering problems between async_mutex and a thread's heap_mutex.
This fixes the following helgrind report.
Thread #1: lock order "0x4325084 before 0x4105328" violated
at 0x40234F7: pthread_mutex_lock (hg_intercepts.c:408)
by 0x40D01EA: scm_i_thread_put_to_sleep (threads.c:1622)
by 0x4078958: scm_make_fluid (fluids.c:114)
by 0x40778D6: scm_init_feature (feature.c:101)
by 0x408C62E: scm_i_init_guile (init.c:464)
by 0x40D1873: scm_i_init_thread_for_guile (threads.c:583)
by 0x40D18B4: scm_i_with_guile_and_parent (threads.c:725)
by 0x40D19BD: scm_with_guile (threads.c:714)
by 0x408C42E: scm_boot_guile (init.c:350)
by 0x8048710: main (guile.c:69)
Required order was established by acquisition of lock at 0x4325084
at 0x40234F7: pthread_mutex_lock (hg_intercepts.c:408)
by 0x40CFFA4: guilify_self_1 (threads.c:468)
by 0x408C58B: scm_i_init_guile (init.c:421)
by 0x40D1873: scm_i_init_thread_for_guile (threads.c:583)
by 0x40D18B4: scm_i_with_guile_and_parent (threads.c:725)
by 0x40D19BD: scm_with_guile (threads.c:714)
by 0x408C42E: scm_boot_guile (init.c:350)
by 0x8048710: main (guile.c:69)
followed by a later acquisition of lock at 0x4105328
at 0x40234F7: pthread_mutex_lock (hg_intercepts.c:408)
by 0x40CFFAC: guilify_self_1 (threads.c:470)
by 0x408C58B: scm_i_init_guile (init.c:421)
by 0x40D1873: scm_i_init_thread_for_guile (threads.c:583)
by 0x40D18B4: scm_i_with_guile_and_parent (threads.c:725)
by 0x40D19BD: scm_with_guile (threads.c:714)
by 0x408C42E: scm_boot_guile (init.c:350)
by 0x8048710: main (guile.c:69)
* threads.c (guilify_self_1): Add self to global thread list _before_
entering Guile mode.
(Reported by Panicz Maciej Godek.)
* test-suite/tests/syncase.test ("@ works with syncase"): New test.
* ice-9/syncase.scm (guile-macro): When a Guile macro transformer
produces a variable, don't pass it through sc-expand.
* ice-9/session.scm (*value-help-handlers*): Define object-documentation
as the default value help handler.
(remove-value-help-handler!, add-name-help-handler!)
(remove-name-help-handler!): Fix docs.
(help, help-doc): Fix so that we try object-documentation through
try-value-help, and we obey the docs regarding what happens with return
values.
* ice-9/session.scm (add-value-help-handler!)
(remove-value-help-handler!, add-name-help-handler!)
(remove-name-help-handler!): New public interfaces, to allow some basic
extensibility of the help interface. Merged in from guile-lib's (scheme
session).
Reported by Roland Haeder. The declaration and definition of
scm_pthread_cond_timedwait were using possibly different types for the
third arg.
* THANKS: Added Roland Haeder.
* libguile/threads.h (scm_pthread_cond_timedwait): Use scm_t_timespec
for third arg rather than struct timespec, for consistency with the
function implementation.
