In buffer.h, prefer inline functions to function-like macros
when either will do. This helps avoid confusion about how
many times an arg is evaluated. On my platform, this patch
improves performance of ‘make compile-always’ by 5.7%.
Also, prefer enum constants to object-like macros
when either will do.
* src/buffer.h (BEG, BEG_BYTE, GAP_BYTES_DFL, GAP_BYTES_MIN)
(MAX_PER_BUFFER_VARS, NONEXISTENT_MODTIME_NSECS)
(UNKNOWN_MODTIME_NSECS, BUFFER_LISP_SIZE, BUFFER_REST_SIZE):
Now enum constants, instead of macros.
(BUFFER_CEILING_OF, BUFFER_FLOOR_OF, BUF_BEG, BUF_BEG_BYTE)
(BUF_BEGV, BUF_BEGV_BYTE, BUF_PT, BUF_PT_BYTE, BUF_ZV)
(BUF_ZV_BYTE, BUF_GPT_ADDR, BUF_Z_ADDR, BUF_GAP_END_ADDR)
(BUF_COMPUTE_UNCHANGED, SET_PT, TEMP_SET_PT, SET_PT_BOTH)
(TEMP_SET_PT_BOTH, BUF_TEMP_SET_PT, SET_BUF_BEGV, SET_BUF_ZV)
(SET_BUF_BEGV_BOTH, SET_BUF_ZV_BOTH, SET_BUF_PT_BOTH)
(BYTE_POS_ADDR, CHAR_POS_ADDR, CHAR_TO_BYTE, BYTE_TO_CHAR)
(PTR_BYTE_POS, FETCH_CHAR, FETCH_CHAR_AS_MULTIBYTE)
(BUF_BYTE_ADDRESS, BUF_CHAR_ADDRESS, BUF_PTR_BYTE_POS)
(BUF_FETCH_CHAR, BUF_FETCH_CHAR_AS_MULTIBYTE, BUF_FETCH_BYTE)
(BUFFER_PVEC_INIT, BUFFER_LIVE_P, BUFFER_HIDDEN_P)
(BUFFER_CHECK_INDIRECTION, OVERLAY_POSITION, PER_BUFFER_VALUE_P)
(SET_PER_BUFFER_VALUE_P, PER_BUFFER_IDX):
Now inline functions instead of macros.
This will simplify future changes. Turn the runtime check
into an eassert, since it’s not needed in production.
* src/buffer.c (last_per_buffer_idx): Now static.
(valid_per_buffer_idx): New function.
* src/buffer.h (PER_BUFFER_VALUE_P, SET_PER_BUFFER_VALUE_P): Use it.
Problem with DUMP_SET_REFERRER mentioned by Pip Cet at end of:
https://lists.gnu.org/archive/html/emacs-devel/2019-07/msg00548.html
* src/pdumper.c (DANGEROUS, EMACS_RELOC_TYPE_BITS)
(EMACS_RELOC_LENGTH_BITS, DUMP_RELOC_TYPE_BITS)
(DUMP_RELOC_ALIGNMENT_BITS, DUMP_RELOC_OFFSET_BITS)
(DUMP_RELOCATION_ALIGNMENT, DUMP_ALIGNMENT)
(WEIGHT_NONE, WEIGHT_NORMAL, WEIGHT_STRONG)
(PDUMPER_MAX_OBJECT_SIZE):
Now a constant, not a macro.
(divide_round_up): Now a function, not a macro DIVIDE_ROUND_UP.
All uses changed.
(enum link_weight_enum, WEIGHT_NONE_VALUE)
(WEIGHT_NORMAL_VALUE, WEIGHT_STRONG_VALUE): Remove.
(struct link_weight): Just use an int.
(dump_set_referrer): New function, replacing DUMP_SET_REFERRER
macro with a different API. All uses changed.
(dump_clear_referrer): Rename from DUMP_CLEAR_REFERRER.
All uses changed.
(DEFINE_FROMLISP_FUNC, DEFINE_TOLISP_FUNC): Remove.
(intmax_t_from_lisp, intmax_t_to_lisp, dump_off_from_lisp)
(dump_off_to_lisp): Define without using macros,
(dump_off_from_lisp): Add an eassert range check.
(DUMP_FIELD_COPY): Simplify.
* src/alloc.c (mark_maybe_pointer):
* src/pdumper.h (pdumper_object_p_precise):
Use pdumper_valid_object_type_p.
* src/pdumper.c (pdumper_find_object_type_impl):
* src/pdumper.h (pdumper_find_object_type):
Return int, not enum Lisp_Type. All callers changed.
* src/pdumper.h (PDUMPER_NO_OBJECT): Do not cast -1 to enum
Lisp_Type; in theory, C18 says this could yield 7, which would
mean PDUMPER_NO_OBJECT == Lisp_Float (!).
(pdumper_valid_object_type_p): New function.
Add watchers for gc-cons-threshold and gc-cons-percentage
that update consing_until_gc accordingly.
Suggested by Eli Zaretskii (Bug#37006#52).
* src/alloc.c (consing_threshold, bump_consing_until_gc)
(watch_gc_cons_threshold, watch_gc_cons_percentage):
New functions.
(garbage_collect_1): Use consing_threshold.
(syms_of_alloc): Arrange to watch gc-cons-threshold and
gc-cons-percentage.
NOTREACHED was designed for traditional lint decades ago,
and _Noreturn now normally subsumes its function.
In the one case in Emacs where NORETURN might help and _Noreturn does
not, check for NOTREACHED instead of assuming it.
* lib-src/etags.c (main):
* src/xterm.c (x_connection_closed):
Remove NOTREACHED after a call to a _Noreturn function, as NOTREACHED
is no longer needed there. Also, one of the NOTREACHEDs was
misplaced, which defeated traditional lint checking anyway.
* lib-src/pop.c (pop_getline): Redo so as to not need NOTREACHED.
* src/emacs.c (main): Use eassume (false) rather than NOTREACHED,
so that running with ENABLE_CHECKING catches any internal error
causing the toplevel Frecursive_edit to return.
* src/gnutls.c (init_gnutls_functions) [WINDOWSNT]: Define and
load gnutls_free by an explicit call to GetProcAddress.
(gnutls_free) [WINDOWSNT]: Define as a macro that dereferences
a function pointer. (Bug#31946)
(Fgnutls_format_certificate): Use make_string_from_bytes
instead of going through an intermediate malloc'ed buffer.
* src/gnutls.c (init_gnutls_functions) [WINDOWSNT]: Define and
load gnutls_free.
(Fgnutls_format_certificate): Use gnutls_free instead of
xfree. This prevents crashes on MS-Windows, since the memory
being released was allocated inside GnuTLS. (Bug#31946)
Problem reported by Roland Winkler (Bug#37153).
Derived from a patch suggested by YAMAMOTO Mitsuharu (Bug#37153#62).
* src/image.c (png_get_valid) [WINDOWSNT]:
Do not dynamically link this function.
(png_get_tRNS) [WINDOWSNT && PNG_tRNS_SUPPORTED]:
Dynamically link this function instead.
(png_load_body): Do not assume that every paletted image supplies
only transparency data. Fix typo in use of transparent_p.
This partially reverts the changes in "extern function cleanup".
* src/coding.c (encode_string_utf_8, decode_string_utf_8): Now
extern again. They should NOT be static, as they are intended
to be used by the likes of json.c, where we need
highly-optimized code for processing UTF-8 strings. E.g.,
decode_string_utf_8 beats make_string_from_utf8 by a factor of
2 to 5 in a large number of scenarios.
* src/data.c (integer_remainder): New function. When the
numerator is a bignum and the denominator is small, this function
uses mpz_tdiv_ui, which should be faster than mpz_tdiv_r.
(Frem, Fmod): Use it.
* src/data.c (integer_mod): Use mpz_tdiv_r not mpz_mod, as that’s
more similar to the fixnum case, is a bit more efficient, and
otherwise the later ‘sgn_r < 0’ code is useless anyway.
* src/data.c (Fmod): Do not signal an error for (mod 1.0 0), for
the same reason (/ 1.0 0) does not signal an error.
* test/src/data-tests.el (data-tests-mod-0): New test.
Most of these functions can be static. A few are unused.
* src/coding.c (encode_string_utf_8, decode_string_utf_8):
Define only if ENABLE_UTF_8_CONVERTER_TEST, as they're
not needed otherwise.
* src/coding.c (encode_string_utf_8, decode_string_utf_8):
* src/data.c (integer_mod):
* src/fns.c (base64_encode_region_1, base64_encode_string_1):
* src/ftfont.c (ftfont_get_fc_charset):
Now static.
* src/sysdep.c (verrprintf): Remove; unused.
* src/gnutls.c (Fgnutls_peer_status): Report :compression and
:encrypt-then-mac only if the underlying GnuTLS library has
the corresponding features. This give the Elisp caller a bit
more information about the peer status.
* lisp/net/nsm.el (nsm-protocol-check--compression):
Don’t worry about compression in newer GnuTLS versions
that do not support compression.
* src/frame.c (Fx_parse_geometry): Pacify the compiler in a
different way, so that the human reader can more easily see
that the initializations are unnecessary.
The recent changes caused the build to fail on Fedora 30 when built
with --enable-gcc-warnings, among other things with diagnostics that
gnutls_compression_get and gnutls_compression_get_name are deprecated
(this started with GnuTLS 3.6). Fix this by refusing to call these
obsolescent and now-dummy functions in GnuTLS 3.6 and later. However,
this is just a temporary workaround to get the build working; a real
fix is needed, as network-stream-tests fail.
* src/gnutls.c (HAVE_GNUTLS_COMPRESSION_GET): New macro.
(gnutls_compression_get, gnutls_compression_get_name):
Define only if HAVE_GNUTLS_COMPRESSION_GET.
(init_gnutls_functions): Load the two functions only if
HAVE_GNUTLS_COMPRESSION_GET.
(emacs_gnutls_certificate_export_pem): Use alloca instead of xmalloc.
(Fgnutls_peer_status): Just return "NULL" if the functions
are deprecated.
(Fgnutls_format_certificate): Fix pointer signedness glitches.
* src/process.c: Fix spacing.
* src/minibuf.c (string_to_object): Signal `invalid-read-syntax'
instead of the generic `error' for "trailing garbage following
expression" (bug#24649).
* src/keymap.c (describe_vector): Remove old workaround for GC bug.
This workaround, introduced in 1993-02-19T05:43:54Z!rms@gnu.org,
has not been needed for some time. Problem reported by Pip Cet in:
https://lists.gnu.org/r/emacs-devel/2019-08/msg00316.html
This GC bug seems to have been fixed, so the check is no longer
needed in production code. From a suggestion by Pip Cet in:
https://lists.gnu.org/r/emacs-devel/2019-08/msg00316.html
* src/alloc.c (SUSPICIOUS_OBJECT_CHECKING) [!ENABLE_CHECKING]:
Do not define.
(find_suspicious_object_in_range, detect_suspicious_free):
Expand to proper dummy expressions if !SUSPICIOUS_OBJECT_CHECKING.
* src/data.c (Ffset): Convert test to an eassert.
Problem reported by Pip Cet in:
https://lists.gnu.org/r/emacs-devel/2019-08/msg00316.html
* src/fns.c (maybe_resize_hash_table): Prefer ASET to gc_aset
where either will do. Simplify appending of Qunbound values.
Put index_size calculation closer to where it’s needed.
(hash_clear): If hash_rehash_needed_p (h), don’t clear the
nonexistent hash vector. Use memclear to speed up clearing.
* src/lisp.h (HASH_TABLE_SIZE): Check that the size is positive,
and tell that to the compiler.
Problem for ‘read’ reported by Pip Cet in:
https://lists.gnu.org/r/emacs-devel/2019-08/msg00316.html
* src/fns.c (Frequire): Protect against circular current-load-list.
* src/lread.c (Fget_load_suffixes):
Protect against circular load-suffixes or load-file-rep-suffixes.
(Fload): Protect against circular loads-in-progress.
(openp): Protect against circular PATH and SUFFIXES.
(build_load_history): Protect against circular load-history or
current-load-list.
(readevalloop_eager_expand_eval): Protect against circular SUBFORMS.
(read1): Protect against circular data.
* test/src/lread-tests.el (lread-circular-hash): New test.
This uses ‘const’ to be better at catching bugs that
mistakenly attempt to modify a bignum value.
Lisp bignums are supposed to be immutable.
* src/alloc.c (make_pure_bignum):
* src/fns.c (sxhash_bignum):
Accept Lisp_Object instead of struct Lisp_Bignum *, as that’s
simpler now. Caller changed.
* src/bignum.h (bignum_val, xbignum_val): New inline functions.
Prefer them to &i->value and XBIGNUM (i)->value, since they
apply ‘const’ to the result.
* src/timefns.c (lisp_to_timespec): Use mpz_t const *
to point to a bignum value.
