Otherwise we can't connect to X server (unix domain socket):
- it is a binary stream (and we get error when we use listen)
- non-blocking io connection doesn't seem to work
Because that needs the DbgHelp library, an additional make
option has been defined for users, who don't want to link to this
library. Some msvc Makefile cleanup has also been done.
Split si_dump_c_backtrace up into two functions:
- _ecl_dump_c_backtrace() (already declared in internal.h, but not
implemented) using only C functions printing to stderr to be used
in case of internal errors
- si_dump_c_backtrace() using lisp functions to be used in the
debugger
Remove broken emulation of backtrace and backtrace_symbols
functions using __builtin_return_address.
- file-length always returns nil
- listen returns correct answer on subsequent reads
- open is non-blocking
- doesn't segfault
note that read still may block (not interruptibly!) if we open stream in IO
mode. process will block until other peer writes something to the stream.
Fixes#242.
- unicode name tables are always compiled in when we have unicode support
- thread local variables support is removed
- profile and rt contribs are deprecated
We should not signal an error if we don't find a dispatch macro
character and *read-suppress* is true. The following example from
the CLHS documentation on *read-suppress* wrongly signaled an
error:
(let ((*read-suppress* t)) (read-from-string "#\garbage"))
Fixes#431.
The current behaviour may technically be ANSI compliant, but it
differs from the behaviour of the native compiler and from many
other Common Lisp implementations (sbcl, ccl, clisp, abcl and cmucl).
The following statement:
(eval-when (:compile-toplevel :load-toplevel :execute)
(print "test"))
resulted in an error, since the compiler would expand the print
statement in a FFI:C-INLINE form, which can't be evaluated at
compile time.
defclass calls indirectly ensure-class-using-class from builtin module which
calls make-instance on metaclass (which could be a forward-referenced
class). That triggers validate-superclass on
(validate-superclass #<forward-ref-class> #<standard-class standard-object>)
which failed. This commit mitigates this situation. Problem was masked
because of *clos-boot* never been set to T (recently fixed).
When we have called something like this
(ext:run-program "true" nil :wait nil :output *standard-output*)
pipe-stream were running in a loop until someone has called
external-process-wait (or external-process-status). This was a hogging the
processor without a reason. Right now unless some other wait is already called
checks for the process status (and magically removes zombies).
Also always call sleep (not thread-yield), because this loop eats a lot of cpu
otherwise.
Some versions of makeinfo are quite strict about input files and
don't allow whitespace between @macros and braces or nodes which
are not referenced in some menu.
:UCS-4LE was not in the list of basic encodings and the keywords
for other encodings were mistakenly interned in lower
case. Reported and fixed by Robert Dodier. Fixes#435.
Avoids redefinition of __ecl_frs_push_result if e.g. two tagbodies
follow each other in the same C block. This redefinition is an
error in C and compilers will not accept the generated code.
Letting the gmp library use uncollectable allocations leads to
leaks, if the bignums are not freed with mpz_clear. Since this
can't be done without a finalizer, we have to use the standard
allocation mechanism. Fixes#433.
Local variables were not declared volatile, so they could be
changed by the setjmp call in ECL_UNWIND_PROTECT_BEGIN. This could
lead to execution of the cleanup statements even though the thread
creation was successfull and thus to segmentation faults.
